arch/s390/kvm/interrupt.c

   1 /*
   2  * handling kvm guest interrupts
   3  *
   4  * Copyright IBM Corp. 2008, 2015
   5  *
   6  * This program is free software; you can redistribute it and/or modify
   7  * it under the terms of the GNU General Public License (version 2 only)
   8  * as published by the Free Software Foundation.
   9  *
  10  *    Author(s): Carsten Otte <cotte@de.ibm.com>
  11  */
  12
  13 #include <linux/interrupt.h>
  14 #include <linux/kvm_host.h>
  15 #include <linux/hrtimer.h>
  16 #include <linux/mmu_context.h>
  17 #include <linux/signal.h>
  18 #include <linux/slab.h>
  19 #include <linux/bitmap.h>
  20 #include <linux/vmalloc.h>
  21 #include <asm/asm-offsets.h>
  22 #include <asm/dis.h>
  23 #include <linux/uaccess.h>
  24 #include <asm/sclp.h>
  25 #include <asm/isc.h>
  26 #include <asm/gmap.h>
  27 #include <asm/switch_to.h>
  28 #include <asm/nmi.h>
  29 #include "kvm-s390.h"
  30 #include "gaccess.h"
  31 #include "trace-s390.h"
  32
  33 #define PFAULT_INIT 0x0600
  34 #define PFAULT_DONE 0x0680
  35 #define VIRTIO_PARAM 0x0d00
  36
  37 /* handle external calls via sigp interpretation facility */
  38 static int sca_ext_call_pending(struct kvm_vcpu *vcpu, int *src_id)
  39 {
  40         int c, scn;
  41
  42         if (!(atomic_read(&vcpu->arch.sie_block->cpuflags) & CPUSTAT_ECALL_PEND))
  43                 return 0;
  44
  45         BUG_ON(!kvm_s390_use_sca_entries());
  46         read_lock(&vcpu->kvm->arch.sca_lock);
  47         if (vcpu->kvm->arch.use_esca) {
  48                 struct esca_block *sca = vcpu->kvm->arch.sca;
  49                 union esca_sigp_ctrl sigp_ctrl =
  50                         sca->cpu[vcpu->vcpu_id].sigp_ctrl;
  51
  52                 c = sigp_ctrl.c;
  53                 scn = sigp_ctrl.scn;
  54         } else {
  55                 struct bsca_block *sca = vcpu->kvm->arch.sca;
  56                 union bsca_sigp_ctrl sigp_ctrl =
  57                         sca->cpu[vcpu->vcpu_id].sigp_ctrl;
  58
  59                 c = sigp_ctrl.c;
  60                 scn = sigp_ctrl.scn;
  61         }
  62         read_unlock(&vcpu->kvm->arch.sca_lock);
  63
  64         if (src_id)
  65                 *src_id = scn;
  66
  67         return c;
  68 }
  69
  70 static int sca_inject_ext_call(struct kvm_vcpu *vcpu, int src_id)
  71 {
  72         int expect, rc;
  73
  74         BUG_ON(!kvm_s390_use_sca_entries());
  75         read_lock(&vcpu->kvm->arch.sca_lock);
  76         if (vcpu->kvm->arch.use_esca) {
  77                 struct esca_block *sca = vcpu->kvm->arch.sca;
  78                 union esca_sigp_ctrl *sigp_ctrl =
  79                         &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
  80                 union esca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
  81
  82                 new_val.scn = src_id;
  83                 new_val.c = 1;
  84                 old_val.c = 0;
  85
  86                 expect = old_val.value;
  87                 rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
  88         } else {
  89                 struct bsca_block *sca = vcpu->kvm->arch.sca;
  90                 union bsca_sigp_ctrl *sigp_ctrl =
  91                         &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
  92                 union bsca_sigp_ctrl new_val = {0}, old_val = *sigp_ctrl;
  93
  94                 new_val.scn = src_id;
  95                 new_val.c = 1;
  96                 old_val.c = 0;
  97
  98                 expect = old_val.value;
  99                 rc = cmpxchg(&sigp_ctrl->value, old_val.value, new_val.value);
 100         }
 101         read_unlock(&vcpu->kvm->arch.sca_lock);
 102
 103         if (rc != expect) {
 104                 /* another external call is pending */
 105                 return -EBUSY;
 106         }
 107         atomic_or(CPUSTAT_ECALL_PEND, &vcpu->arch.sie_block->cpuflags);
 108         return 0;
 109 }
 110
 111 static void sca_clear_ext_call(struct kvm_vcpu *vcpu)
 112 {
 113         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 114         int rc, expect;
 115
 116         if (!kvm_s390_use_sca_entries())
 117                 return;
 118         atomic_andnot(CPUSTAT_ECALL_PEND, li->cpuflags);
 119         read_lock(&vcpu->kvm->arch.sca_lock);
 120         if (vcpu->kvm->arch.use_esca) {
 121                 struct esca_block *sca = vcpu->kvm->arch.sca;
 122                 union esca_sigp_ctrl *sigp_ctrl =
 123                         &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
 124                 union esca_sigp_ctrl old = *sigp_ctrl;
 125
 126                 expect = old.value;
 127                 rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
 128         } else {
 129                 struct bsca_block *sca = vcpu->kvm->arch.sca;
 130                 union bsca_sigp_ctrl *sigp_ctrl =
 131                         &(sca->cpu[vcpu->vcpu_id].sigp_ctrl);
 132                 union bsca_sigp_ctrl old = *sigp_ctrl;
 133
 134                 expect = old.value;
 135                 rc = cmpxchg(&sigp_ctrl->value, old.value, 0);
 136         }
 137         read_unlock(&vcpu->kvm->arch.sca_lock);
 138         WARN_ON(rc != expect); /* cannot clear? */
 139 }
 140
 141 int psw_extint_disabled(struct kvm_vcpu *vcpu)
 142 {
 143         return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_EXT);
 144 }
 145
 146 static int psw_ioint_disabled(struct kvm_vcpu *vcpu)
 147 {
 148         return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_IO);
 149 }
 150
 151 static int psw_mchk_disabled(struct kvm_vcpu *vcpu)
 152 {
 153         return !(vcpu->arch.sie_block->gpsw.mask & PSW_MASK_MCHECK);
 154 }
 155
 156 static int psw_interrupts_disabled(struct kvm_vcpu *vcpu)
 157 {
 158         return psw_extint_disabled(vcpu) &&
 159                psw_ioint_disabled(vcpu) &&
 160                psw_mchk_disabled(vcpu);
 161 }
 162
 163 static int ckc_interrupts_enabled(struct kvm_vcpu *vcpu)
 164 {
 165         if (psw_extint_disabled(vcpu) ||
 166             !(vcpu->arch.sie_block->gcr[0] & 0x800ul))
 167                 return 0;
 168         if (guestdbg_enabled(vcpu) && guestdbg_sstep_enabled(vcpu))
 169                 /* No timer interrupts when single stepping */
 170                 return 0;
 171         return 1;
 172 }
 173
 174 static int ckc_irq_pending(struct kvm_vcpu *vcpu)
 175 {
 176         const u64 now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
 177         const u64 ckc = vcpu->arch.sie_block->ckc;
 178
 179         if (vcpu->arch.sie_block->gcr[0] & 0x0020000000000000ul) {
 180                 if ((s64)ckc >= (s64)now)
 181                         return 0;
 182         } else if (ckc >= now) {
 183                 return 0;
 184         }
 185         return ckc_interrupts_enabled(vcpu);
 186 }
 187
 188 static int cpu_timer_interrupts_enabled(struct kvm_vcpu *vcpu)
 189 {
 190         return !psw_extint_disabled(vcpu) &&
 191                (vcpu->arch.sie_block->gcr[0] & 0x400ul);
 192 }
 193
 194 static int cpu_timer_irq_pending(struct kvm_vcpu *vcpu)
 195 {
 196         if (!cpu_timer_interrupts_enabled(vcpu))
 197                 return 0;
 198         return kvm_s390_get_cpu_timer(vcpu) >> 63;
 199 }
 200
 201 static inline int is_ioirq(unsigned long irq_type)
 202 {
 203         return ((irq_type >= IRQ_PEND_IO_ISC_0) &&
 204                 (irq_type <= IRQ_PEND_IO_ISC_7));
 205 }
 206
 207 static uint64_t isc_to_isc_bits(int isc)
 208 {
 209         return (0x80 >> isc) << 24;
 210 }
 211
 212 static inline u8 int_word_to_isc(u32 int_word)
 213 {
 214         return (int_word & 0x38000000) >> 27;
 215 }
 216
 217 static inline unsigned long pending_irqs(struct kvm_vcpu *vcpu)
 218 {
 219         return vcpu->kvm->arch.float_int.pending_irqs |
 220                vcpu->arch.local_int.pending_irqs;
 221 }
 222
 223 static unsigned long disable_iscs(struct kvm_vcpu *vcpu,
 224                                    unsigned long active_mask)
 225 {
 226         int i;
 227
 228         for (i = 0; i <= MAX_ISC; i++)
 229                 if (!(vcpu->arch.sie_block->gcr[6] & isc_to_isc_bits(i)))
 230                         active_mask &= ~(1UL << (IRQ_PEND_IO_ISC_0 + i));
 231
 232         return active_mask;
 233 }
 234
 235 static unsigned long deliverable_irqs(struct kvm_vcpu *vcpu)
 236 {
 237         unsigned long active_mask;
 238
 239         active_mask = pending_irqs(vcpu);
 240         if (!active_mask)
 241                 return 0;
 242
 243         if (psw_extint_disabled(vcpu))
 244                 active_mask &= ~IRQ_PEND_EXT_MASK;
 245         if (psw_ioint_disabled(vcpu))
 246                 active_mask &= ~IRQ_PEND_IO_MASK;
 247         else
 248                 active_mask = disable_iscs(vcpu, active_mask);
 249         if (!(vcpu->arch.sie_block->gcr[0] & 0x2000ul))
 250                 __clear_bit(IRQ_PEND_EXT_EXTERNAL, &active_mask);
 251         if (!(vcpu->arch.sie_block->gcr[0] & 0x4000ul))
 252                 __clear_bit(IRQ_PEND_EXT_EMERGENCY, &active_mask);
 253         if (!(vcpu->arch.sie_block->gcr[0] & 0x800ul))
 254                 __clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &active_mask);
 255         if (!(vcpu->arch.sie_block->gcr[0] & 0x400ul))
 256                 __clear_bit(IRQ_PEND_EXT_CPU_TIMER, &active_mask);
 257         if (!(vcpu->arch.sie_block->gcr[0] & 0x200ul))
 258                 __clear_bit(IRQ_PEND_EXT_SERVICE, &active_mask);
 259         if (psw_mchk_disabled(vcpu))
 260                 active_mask &= ~IRQ_PEND_MCHK_MASK;
 261         /*
 262          * Check both floating and local interrupt's cr14 because
 263          * bit IRQ_PEND_MCHK_REP could be set in both cases.
 264          */
 265         if (!(vcpu->arch.sie_block->gcr[14] &
 266            (vcpu->kvm->arch.float_int.mchk.cr14 |
 267            vcpu->arch.local_int.irq.mchk.cr14)))
 268                 __clear_bit(IRQ_PEND_MCHK_REP, &active_mask);
 269
 270         /*
 271          * STOP irqs will never be actively delivered. They are triggered via
 272          * intercept requests and cleared when the stop intercept is performed.
 273          */
 274         __clear_bit(IRQ_PEND_SIGP_STOP, &active_mask);
 275
 276         return active_mask;
 277 }
 278
 279 static void __set_cpu_idle(struct kvm_vcpu *vcpu)
 280 {
 281         atomic_or(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
 282         set_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
 283 }
 284
 285 static void __unset_cpu_idle(struct kvm_vcpu *vcpu)
 286 {
 287         atomic_andnot(CPUSTAT_WAIT, &vcpu->arch.sie_block->cpuflags);
 288         clear_bit(vcpu->vcpu_id, vcpu->arch.local_int.float_int->idle_mask);
 289 }
 290
 291 static void __reset_intercept_indicators(struct kvm_vcpu *vcpu)
 292 {
 293         atomic_andnot(CPUSTAT_IO_INT | CPUSTAT_EXT_INT | CPUSTAT_STOP_INT,
 294                     &vcpu->arch.sie_block->cpuflags);
 295         vcpu->arch.sie_block->lctl = 0x0000;
 296         vcpu->arch.sie_block->ictl &= ~(ICTL_LPSW | ICTL_STCTL | ICTL_PINT);
 297
 298         if (guestdbg_enabled(vcpu)) {
 299                 vcpu->arch.sie_block->lctl |= (LCTL_CR0 | LCTL_CR9 |
 300                                                LCTL_CR10 | LCTL_CR11);
 301                 vcpu->arch.sie_block->ictl |= (ICTL_STCTL | ICTL_PINT);
 302         }
 303 }
 304
 305 static void __set_cpuflag(struct kvm_vcpu *vcpu, u32 flag)
 306 {
 307         atomic_or(flag, &vcpu->arch.sie_block->cpuflags);
 308 }
 309
 310 static void set_intercept_indicators_io(struct kvm_vcpu *vcpu)
 311 {
 312         if (!(pending_irqs(vcpu) & IRQ_PEND_IO_MASK))
 313                 return;
 314         else if (psw_ioint_disabled(vcpu))
 315                 __set_cpuflag(vcpu, CPUSTAT_IO_INT);
 316         else
 317                 vcpu->arch.sie_block->lctl |= LCTL_CR6;
 318 }
 319
 320 static void set_intercept_indicators_ext(struct kvm_vcpu *vcpu)
 321 {
 322         if (!(pending_irqs(vcpu) & IRQ_PEND_EXT_MASK))
 323                 return;
 324         if (psw_extint_disabled(vcpu))
 325                 __set_cpuflag(vcpu, CPUSTAT_EXT_INT);
 326         else
 327                 vcpu->arch.sie_block->lctl |= LCTL_CR0;
 328 }
 329
 330 static void set_intercept_indicators_mchk(struct kvm_vcpu *vcpu)
 331 {
 332         if (!(pending_irqs(vcpu) & IRQ_PEND_MCHK_MASK))
 333                 return;
 334         if (psw_mchk_disabled(vcpu))
 335                 vcpu->arch.sie_block->ictl |= ICTL_LPSW;
 336         else
 337                 vcpu->arch.sie_block->lctl |= LCTL_CR14;
 338 }
 339
 340 static void set_intercept_indicators_stop(struct kvm_vcpu *vcpu)
 341 {
 342         if (kvm_s390_is_stop_irq_pending(vcpu))
 343                 __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
 344 }
 345
 346 /* Set interception request for non-deliverable interrupts */
 347 static void set_intercept_indicators(struct kvm_vcpu *vcpu)
 348 {
 349         set_intercept_indicators_io(vcpu);
 350         set_intercept_indicators_ext(vcpu);
 351         set_intercept_indicators_mchk(vcpu);
 352         set_intercept_indicators_stop(vcpu);
 353 }
 354
 355 static int __must_check __deliver_cpu_timer(struct kvm_vcpu *vcpu)
 356 {
 357         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 358         int rc;
 359
 360         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
 361                                          0, 0);
 362
 363         rc  = put_guest_lc(vcpu, EXT_IRQ_CPU_TIMER,
 364                            (u16 *)__LC_EXT_INT_CODE);
 365         rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
 366         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 367                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 368         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 369                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 370         clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
 371         return rc ? -EFAULT : 0;
 372 }
 373
 374 static int __must_check __deliver_ckc(struct kvm_vcpu *vcpu)
 375 {
 376         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 377         int rc;
 378
 379         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
 380                                          0, 0);
 381
 382         rc  = put_guest_lc(vcpu, EXT_IRQ_CLK_COMP,
 383                            (u16 __user *)__LC_EXT_INT_CODE);
 384         rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
 385         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 386                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 387         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 388                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 389         clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
 390         return rc ? -EFAULT : 0;
 391 }
 392
 393 static int __must_check __deliver_pfault_init(struct kvm_vcpu *vcpu)
 394 {
 395         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 396         struct kvm_s390_ext_info ext;
 397         int rc;
 398
 399         spin_lock(&li->lock);
 400         ext = li->irq.ext;
 401         clear_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
 402         li->irq.ext.ext_params2 = 0;
 403         spin_unlock(&li->lock);
 404
 405         VCPU_EVENT(vcpu, 4, "deliver: pfault init token 0x%llx",
 406                    ext.ext_params2);
 407         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 408                                          KVM_S390_INT_PFAULT_INIT,
 409                                          0, ext.ext_params2);
 410
 411         rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE, (u16 *) __LC_EXT_INT_CODE);
 412         rc |= put_guest_lc(vcpu, PFAULT_INIT, (u16 *) __LC_EXT_CPU_ADDR);
 413         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 414                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 415         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 416                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 417         rc |= put_guest_lc(vcpu, ext.ext_params2, (u64 *) __LC_EXT_PARAMS2);
 418         return rc ? -EFAULT : 0;
 419 }
 420
 421 static int __write_machine_check(struct kvm_vcpu *vcpu,
 422                                  struct kvm_s390_mchk_info *mchk)
 423 {
 424         unsigned long ext_sa_addr;
 425         unsigned long lc;
 426         freg_t fprs[NUM_FPRS];
 427         union mci mci;
 428         int rc;
 429
 430         mci.val = mchk->mcic;
 431         /* take care of lazy register loading */
 432         save_fpu_regs();
 433         save_access_regs(vcpu->run->s.regs.acrs);
 434         if (MACHINE_HAS_GS && vcpu->arch.gs_enabled)
 435                 save_gs_cb(current->thread.gs_cb);
 436
 437         /* Extended save area */
 438         rc = read_guest_lc(vcpu, __LC_MCESAD, &ext_sa_addr,
 439                            sizeof(unsigned long));
 440         /* Only bits 0 through 63-LC are used for address formation */
 441         lc = ext_sa_addr & MCESA_LC_MASK;
 442         if (test_kvm_facility(vcpu->kvm, 133)) {
 443                 switch (lc) {
 444                 case 0:
 445                 case 10:
 446                         ext_sa_addr &= ~0x3ffUL;
 447                         break;
 448                 case 11:
 449                         ext_sa_addr &= ~0x7ffUL;
 450                         break;
 451                 case 12:
 452                         ext_sa_addr &= ~0xfffUL;
 453                         break;
 454                 default:
 455                         ext_sa_addr = 0;
 456                         break;
 457                 }
 458         } else {
 459                 ext_sa_addr &= ~0x3ffUL;
 460         }
 461
 462         if (!rc && mci.vr && ext_sa_addr && test_kvm_facility(vcpu->kvm, 129)) {
 463                 if (write_guest_abs(vcpu, ext_sa_addr, vcpu->run->s.regs.vrs,
 464                                     512))
 465                         mci.vr = 0;
 466         } else {
 467                 mci.vr = 0;
 468         }
 469         if (!rc && mci.gs && ext_sa_addr && test_kvm_facility(vcpu->kvm, 133)
 470             && (lc == 11 || lc == 12)) {
 471                 if (write_guest_abs(vcpu, ext_sa_addr + 1024,
 472                                     &vcpu->run->s.regs.gscb, 32))
 473                         mci.gs = 0;
 474         } else {
 475                 mci.gs = 0;
 476         }
 477
 478         /* General interruption information */
 479         rc |= put_guest_lc(vcpu, 1, (u8 __user *) __LC_AR_MODE_ID);
 480         rc |= write_guest_lc(vcpu, __LC_MCK_OLD_PSW,
 481                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 482         rc |= read_guest_lc(vcpu, __LC_MCK_NEW_PSW,
 483                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 484         rc |= put_guest_lc(vcpu, mci.val, (u64 __user *) __LC_MCCK_CODE);
 485
 486         /* Register-save areas */
 487         if (MACHINE_HAS_VX) {
 488                 convert_vx_to_fp(fprs, (__vector128 *) vcpu->run->s.regs.vrs);
 489                 rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA, fprs, 128);
 490         } else {
 491                 rc |= write_guest_lc(vcpu, __LC_FPREGS_SAVE_AREA,
 492                                      vcpu->run->s.regs.fprs, 128);
 493         }
 494         rc |= write_guest_lc(vcpu, __LC_GPREGS_SAVE_AREA,
 495                              vcpu->run->s.regs.gprs, 128);
 496         rc |= put_guest_lc(vcpu, current->thread.fpu.fpc,
 497                            (u32 __user *) __LC_FP_CREG_SAVE_AREA);
 498         rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->todpr,
 499                            (u32 __user *) __LC_TOD_PROGREG_SAVE_AREA);
 500         rc |= put_guest_lc(vcpu, kvm_s390_get_cpu_timer(vcpu),
 501                            (u64 __user *) __LC_CPU_TIMER_SAVE_AREA);
 502         rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->ckc >> 8,
 503                            (u64 __user *) __LC_CLOCK_COMP_SAVE_AREA);
 504         rc |= write_guest_lc(vcpu, __LC_AREGS_SAVE_AREA,
 505                              &vcpu->run->s.regs.acrs, 64);
 506         rc |= write_guest_lc(vcpu, __LC_CREGS_SAVE_AREA,
 507                              &vcpu->arch.sie_block->gcr, 128);
 508
 509         /* Extended interruption information */
 510         rc |= put_guest_lc(vcpu, mchk->ext_damage_code,
 511                            (u32 __user *) __LC_EXT_DAMAGE_CODE);
 512         rc |= put_guest_lc(vcpu, mchk->failing_storage_address,
 513                            (u64 __user *) __LC_MCCK_FAIL_STOR_ADDR);
 514         rc |= write_guest_lc(vcpu, __LC_PSW_SAVE_AREA, &mchk->fixed_logout,
 515                              sizeof(mchk->fixed_logout));
 516         return rc ? -EFAULT : 0;
 517 }
 518
 519 static int __must_check __deliver_machine_check(struct kvm_vcpu *vcpu)
 520 {
 521         struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
 522         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 523         struct kvm_s390_mchk_info mchk = {};
 524         int deliver = 0;
 525         int rc = 0;
 526
 527         spin_lock(&fi->lock);
 528         spin_lock(&li->lock);
 529         if (test_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs) ||
 530             test_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs)) {
 531                 /*
 532                  * If there was an exigent machine check pending, then any
 533                  * repressible machine checks that might have been pending
 534                  * are indicated along with it, so always clear bits for
 535                  * repressible and exigent interrupts
 536                  */
 537                 mchk = li->irq.mchk;
 538                 clear_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
 539                 clear_bit(IRQ_PEND_MCHK_REP, &li->pending_irqs);
 540                 memset(&li->irq.mchk, 0, sizeof(mchk));
 541                 deliver = 1;
 542         }
 543         /*
 544          * We indicate floating repressible conditions along with
 545          * other pending conditions. Channel Report Pending and Channel
 546          * Subsystem damage are the only two and and are indicated by
 547          * bits in mcic and masked in cr14.
 548          */
 549         if (test_and_clear_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
 550                 mchk.mcic |= fi->mchk.mcic;
 551                 mchk.cr14 |= fi->mchk.cr14;
 552                 memset(&fi->mchk, 0, sizeof(mchk));
 553                 deliver = 1;
 554         }
 555         spin_unlock(&li->lock);
 556         spin_unlock(&fi->lock);
 557
 558         if (deliver) {
 559                 VCPU_EVENT(vcpu, 3, "deliver: machine check mcic 0x%llx",
 560                            mchk.mcic);
 561                 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 562                                                  KVM_S390_MCHK,
 563                                                  mchk.cr14, mchk.mcic);
 564                 rc = __write_machine_check(vcpu, &mchk);
 565         }
 566         return rc;
 567 }
 568
 569 static int __must_check __deliver_restart(struct kvm_vcpu *vcpu)
 570 {
 571         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 572         int rc;
 573
 574         VCPU_EVENT(vcpu, 3, "%s", "deliver: cpu restart");
 575         vcpu->stat.deliver_restart_signal++;
 576         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
 577
 578         rc  = write_guest_lc(vcpu,
 579                              offsetof(struct lowcore, restart_old_psw),
 580                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 581         rc |= read_guest_lc(vcpu, offsetof(struct lowcore, restart_psw),
 582                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 583         clear_bit(IRQ_PEND_RESTART, &li->pending_irqs);
 584         return rc ? -EFAULT : 0;
 585 }
 586
 587 static int __must_check __deliver_set_prefix(struct kvm_vcpu *vcpu)
 588 {
 589         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 590         struct kvm_s390_prefix_info prefix;
 591
 592         spin_lock(&li->lock);
 593         prefix = li->irq.prefix;
 594         li->irq.prefix.address = 0;
 595         clear_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
 596         spin_unlock(&li->lock);
 597
 598         vcpu->stat.deliver_prefix_signal++;
 599         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 600                                          KVM_S390_SIGP_SET_PREFIX,
 601                                          prefix.address, 0);
 602
 603         kvm_s390_set_prefix(vcpu, prefix.address);
 604         return 0;
 605 }
 606
 607 static int __must_check __deliver_emergency_signal(struct kvm_vcpu *vcpu)
 608 {
 609         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 610         int rc;
 611         int cpu_addr;
 612
 613         spin_lock(&li->lock);
 614         cpu_addr = find_first_bit(li->sigp_emerg_pending, KVM_MAX_VCPUS);
 615         clear_bit(cpu_addr, li->sigp_emerg_pending);
 616         if (bitmap_empty(li->sigp_emerg_pending, KVM_MAX_VCPUS))
 617                 clear_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
 618         spin_unlock(&li->lock);
 619
 620         VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp emerg");
 621         vcpu->stat.deliver_emergency_signal++;
 622         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
 623                                          cpu_addr, 0);
 624
 625         rc  = put_guest_lc(vcpu, EXT_IRQ_EMERGENCY_SIG,
 626                            (u16 *)__LC_EXT_INT_CODE);
 627         rc |= put_guest_lc(vcpu, cpu_addr, (u16 *)__LC_EXT_CPU_ADDR);
 628         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 629                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 630         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 631                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 632         return rc ? -EFAULT : 0;
 633 }
 634
 635 static int __must_check __deliver_external_call(struct kvm_vcpu *vcpu)
 636 {
 637         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 638         struct kvm_s390_extcall_info extcall;
 639         int rc;
 640
 641         spin_lock(&li->lock);
 642         extcall = li->irq.extcall;
 643         li->irq.extcall.code = 0;
 644         clear_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
 645         spin_unlock(&li->lock);
 646
 647         VCPU_EVENT(vcpu, 4, "%s", "deliver: sigp ext call");
 648         vcpu->stat.deliver_external_call++;
 649         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 650                                          KVM_S390_INT_EXTERNAL_CALL,
 651                                          extcall.code, 0);
 652
 653         rc  = put_guest_lc(vcpu, EXT_IRQ_EXTERNAL_CALL,
 654                            (u16 *)__LC_EXT_INT_CODE);
 655         rc |= put_guest_lc(vcpu, extcall.code, (u16 *)__LC_EXT_CPU_ADDR);
 656         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 657                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 658         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW, &vcpu->arch.sie_block->gpsw,
 659                             sizeof(psw_t));
 660         return rc ? -EFAULT : 0;
 661 }
 662
 663 static int __must_check __deliver_prog(struct kvm_vcpu *vcpu)
 664 {
 665         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 666         struct kvm_s390_pgm_info pgm_info;
 667         int rc = 0, nullifying = false;
 668         u16 ilen;
 669
 670         spin_lock(&li->lock);
 671         pgm_info = li->irq.pgm;
 672         clear_bit(IRQ_PEND_PROG, &li->pending_irqs);
 673         memset(&li->irq.pgm, 0, sizeof(pgm_info));
 674         spin_unlock(&li->lock);
 675
 676         ilen = pgm_info.flags & KVM_S390_PGM_FLAGS_ILC_MASK;
 677         VCPU_EVENT(vcpu, 3, "deliver: program irq code 0x%x, ilen:%d",
 678                    pgm_info.code, ilen);
 679         vcpu->stat.deliver_program_int++;
 680         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
 681                                          pgm_info.code, 0);
 682
 683         switch (pgm_info.code & ~PGM_PER) {
 684         case PGM_AFX_TRANSLATION:
 685         case PGM_ASX_TRANSLATION:
 686         case PGM_EX_TRANSLATION:
 687         case PGM_LFX_TRANSLATION:
 688         case PGM_LSTE_SEQUENCE:
 689         case PGM_LSX_TRANSLATION:
 690         case PGM_LX_TRANSLATION:
 691         case PGM_PRIMARY_AUTHORITY:
 692         case PGM_SECONDARY_AUTHORITY:
 693                 nullifying = true;
 694                 /* fall through */
 695         case PGM_SPACE_SWITCH:
 696                 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
 697                                   (u64 *)__LC_TRANS_EXC_CODE);
 698                 break;
 699         case PGM_ALEN_TRANSLATION:
 700         case PGM_ALE_SEQUENCE:
 701         case PGM_ASTE_INSTANCE:
 702         case PGM_ASTE_SEQUENCE:
 703         case PGM_ASTE_VALIDITY:
 704         case PGM_EXTENDED_AUTHORITY:
 705                 rc = put_guest_lc(vcpu, pgm_info.exc_access_id,
 706                                   (u8 *)__LC_EXC_ACCESS_ID);
 707                 nullifying = true;
 708                 break;
 709         case PGM_ASCE_TYPE:
 710         case PGM_PAGE_TRANSLATION:
 711         case PGM_REGION_FIRST_TRANS:
 712         case PGM_REGION_SECOND_TRANS:
 713         case PGM_REGION_THIRD_TRANS:
 714         case PGM_SEGMENT_TRANSLATION:
 715                 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
 716                                   (u64 *)__LC_TRANS_EXC_CODE);
 717                 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
 718                                    (u8 *)__LC_EXC_ACCESS_ID);
 719                 rc |= put_guest_lc(vcpu, pgm_info.op_access_id,
 720                                    (u8 *)__LC_OP_ACCESS_ID);
 721                 nullifying = true;
 722                 break;
 723         case PGM_MONITOR:
 724                 rc = put_guest_lc(vcpu, pgm_info.mon_class_nr,
 725                                   (u16 *)__LC_MON_CLASS_NR);
 726                 rc |= put_guest_lc(vcpu, pgm_info.mon_code,
 727                                    (u64 *)__LC_MON_CODE);
 728                 break;
 729         case PGM_VECTOR_PROCESSING:
 730         case PGM_DATA:
 731                 rc = put_guest_lc(vcpu, pgm_info.data_exc_code,
 732                                   (u32 *)__LC_DATA_EXC_CODE);
 733                 break;
 734         case PGM_PROTECTION:
 735                 rc = put_guest_lc(vcpu, pgm_info.trans_exc_code,
 736                                   (u64 *)__LC_TRANS_EXC_CODE);
 737                 rc |= put_guest_lc(vcpu, pgm_info.exc_access_id,
 738                                    (u8 *)__LC_EXC_ACCESS_ID);
 739                 break;
 740         case PGM_STACK_FULL:
 741         case PGM_STACK_EMPTY:
 742         case PGM_STACK_SPECIFICATION:
 743         case PGM_STACK_TYPE:
 744         case PGM_STACK_OPERATION:
 745         case PGM_TRACE_TABEL:
 746         case PGM_CRYPTO_OPERATION:
 747                 nullifying = true;
 748                 break;
 749         }
 750
 751         if (pgm_info.code & PGM_PER) {
 752                 rc |= put_guest_lc(vcpu, pgm_info.per_code,
 753                                    (u8 *) __LC_PER_CODE);
 754                 rc |= put_guest_lc(vcpu, pgm_info.per_atmid,
 755                                    (u8 *)__LC_PER_ATMID);
 756                 rc |= put_guest_lc(vcpu, pgm_info.per_address,
 757                                    (u64 *) __LC_PER_ADDRESS);
 758                 rc |= put_guest_lc(vcpu, pgm_info.per_access_id,
 759                                    (u8 *) __LC_PER_ACCESS_ID);
 760         }
 761
 762         if (nullifying && !(pgm_info.flags & KVM_S390_PGM_FLAGS_NO_REWIND))
 763                 kvm_s390_rewind_psw(vcpu, ilen);
 764
 765         /* bit 1+2 of the target are the ilc, so we can directly use ilen */
 766         rc |= put_guest_lc(vcpu, ilen, (u16 *) __LC_PGM_ILC);
 767         rc |= put_guest_lc(vcpu, vcpu->arch.sie_block->gbea,
 768                                  (u64 *) __LC_LAST_BREAK);
 769         rc |= put_guest_lc(vcpu, pgm_info.code,
 770                            (u16 *)__LC_PGM_INT_CODE);
 771         rc |= write_guest_lc(vcpu, __LC_PGM_OLD_PSW,
 772                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 773         rc |= read_guest_lc(vcpu, __LC_PGM_NEW_PSW,
 774                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 775         return rc ? -EFAULT : 0;
 776 }
 777
 778 static int __must_check __deliver_service(struct kvm_vcpu *vcpu)
 779 {
 780         struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
 781         struct kvm_s390_ext_info ext;
 782         int rc = 0;
 783
 784         spin_lock(&fi->lock);
 785         if (!(test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs))) {
 786                 spin_unlock(&fi->lock);
 787                 return 0;
 788         }
 789         ext = fi->srv_signal;
 790         memset(&fi->srv_signal, 0, sizeof(ext));
 791         clear_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
 792         spin_unlock(&fi->lock);
 793
 794         VCPU_EVENT(vcpu, 4, "deliver: sclp parameter 0x%x",
 795                    ext.ext_params);
 796         vcpu->stat.deliver_service_signal++;
 797         trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id, KVM_S390_INT_SERVICE,
 798                                          ext.ext_params, 0);
 799
 800         rc  = put_guest_lc(vcpu, EXT_IRQ_SERVICE_SIG, (u16 *)__LC_EXT_INT_CODE);
 801         rc |= put_guest_lc(vcpu, 0, (u16 *)__LC_EXT_CPU_ADDR);
 802         rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 803                              &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 804         rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 805                             &vcpu->arch.sie_block->gpsw, sizeof(psw_t));
 806         rc |= put_guest_lc(vcpu, ext.ext_params,
 807                            (u32 *)__LC_EXT_PARAMS);
 808
 809         return rc ? -EFAULT : 0;
 810 }
 811
 812 static int __must_check __deliver_pfault_done(struct kvm_vcpu *vcpu)
 813 {
 814         struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
 815         struct kvm_s390_interrupt_info *inti;
 816         int rc = 0;
 817
 818         spin_lock(&fi->lock);
 819         inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_PFAULT],
 820                                         struct kvm_s390_interrupt_info,
 821                                         list);
 822         if (inti) {
 823                 list_del(&inti->list);
 824                 fi->counters[FIRQ_CNTR_PFAULT] -= 1;
 825         }
 826         if (list_empty(&fi->lists[FIRQ_LIST_PFAULT]))
 827                 clear_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
 828         spin_unlock(&fi->lock);
 829
 830         if (inti) {
 831                 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 832                                                  KVM_S390_INT_PFAULT_DONE, 0,
 833                                                  inti->ext.ext_params2);
 834                 VCPU_EVENT(vcpu, 4, "deliver: pfault done token 0x%llx",
 835                            inti->ext.ext_params2);
 836
 837                 rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
 838                                 (u16 *)__LC_EXT_INT_CODE);
 839                 rc |= put_guest_lc(vcpu, PFAULT_DONE,
 840                                 (u16 *)__LC_EXT_CPU_ADDR);
 841                 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 842                                 &vcpu->arch.sie_block->gpsw,
 843                                 sizeof(psw_t));
 844                 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 845                                 &vcpu->arch.sie_block->gpsw,
 846                                 sizeof(psw_t));
 847                 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
 848                                 (u64 *)__LC_EXT_PARAMS2);
 849                 kfree(inti);
 850         }
 851         return rc ? -EFAULT : 0;
 852 }
 853
 854 static int __must_check __deliver_virtio(struct kvm_vcpu *vcpu)
 855 {
 856         struct kvm_s390_float_interrupt *fi = &vcpu->kvm->arch.float_int;
 857         struct kvm_s390_interrupt_info *inti;
 858         int rc = 0;
 859
 860         spin_lock(&fi->lock);
 861         inti = list_first_entry_or_null(&fi->lists[FIRQ_LIST_VIRTIO],
 862                                         struct kvm_s390_interrupt_info,
 863                                         list);
 864         if (inti) {
 865                 VCPU_EVENT(vcpu, 4,
 866                            "deliver: virtio parm: 0x%x,parm64: 0x%llx",
 867                            inti->ext.ext_params, inti->ext.ext_params2);
 868                 vcpu->stat.deliver_virtio_interrupt++;
 869                 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 870                                 inti->type,
 871                                 inti->ext.ext_params,
 872                                 inti->ext.ext_params2);
 873                 list_del(&inti->list);
 874                 fi->counters[FIRQ_CNTR_VIRTIO] -= 1;
 875         }
 876         if (list_empty(&fi->lists[FIRQ_LIST_VIRTIO]))
 877                 clear_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
 878         spin_unlock(&fi->lock);
 879
 880         if (inti) {
 881                 rc  = put_guest_lc(vcpu, EXT_IRQ_CP_SERVICE,
 882                                 (u16 *)__LC_EXT_INT_CODE);
 883                 rc |= put_guest_lc(vcpu, VIRTIO_PARAM,
 884                                 (u16 *)__LC_EXT_CPU_ADDR);
 885                 rc |= write_guest_lc(vcpu, __LC_EXT_OLD_PSW,
 886                                 &vcpu->arch.sie_block->gpsw,
 887                                 sizeof(psw_t));
 888                 rc |= read_guest_lc(vcpu, __LC_EXT_NEW_PSW,
 889                                 &vcpu->arch.sie_block->gpsw,
 890                                 sizeof(psw_t));
 891                 rc |= put_guest_lc(vcpu, inti->ext.ext_params,
 892                                 (u32 *)__LC_EXT_PARAMS);
 893                 rc |= put_guest_lc(vcpu, inti->ext.ext_params2,
 894                                 (u64 *)__LC_EXT_PARAMS2);
 895                 kfree(inti);
 896         }
 897         return rc ? -EFAULT : 0;
 898 }
 899
 900 static int __must_check __deliver_io(struct kvm_vcpu *vcpu,
 901                                      unsigned long irq_type)
 902 {
 903         struct list_head *isc_list;
 904         struct kvm_s390_float_interrupt *fi;
 905         struct kvm_s390_interrupt_info *inti = NULL;
 906         int rc = 0;
 907
 908         fi = &vcpu->kvm->arch.float_int;
 909
 910         spin_lock(&fi->lock);
 911         isc_list = &fi->lists[irq_type - IRQ_PEND_IO_ISC_0];
 912         inti = list_first_entry_or_null(isc_list,
 913                                         struct kvm_s390_interrupt_info,
 914                                         list);
 915         if (inti) {
 916                 if (inti->type & KVM_S390_INT_IO_AI_MASK)
 917                         VCPU_EVENT(vcpu, 4, "%s", "deliver: I/O (AI)");
 918                 else
 919                         VCPU_EVENT(vcpu, 4, "deliver: I/O %x ss %x schid %04x",
 920                         inti->io.subchannel_id >> 8,
 921                         inti->io.subchannel_id >> 1 & 0x3,
 922                         inti->io.subchannel_nr);
 923
 924                 vcpu->stat.deliver_io_int++;
 925                 trace_kvm_s390_deliver_interrupt(vcpu->vcpu_id,
 926                                 inti->type,
 927                                 ((__u32)inti->io.subchannel_id << 16) |
 928                                 inti->io.subchannel_nr,
 929                                 ((__u64)inti->io.io_int_parm << 32) |
 930                                 inti->io.io_int_word);
 931                 list_del(&inti->list);
 932                 fi->counters[FIRQ_CNTR_IO] -= 1;
 933         }
 934         if (list_empty(isc_list))
 935                 clear_bit(irq_type, &fi->pending_irqs);
 936         spin_unlock(&fi->lock);
 937
 938         if (inti) {
 939                 rc  = put_guest_lc(vcpu, inti->io.subchannel_id,
 940                                 (u16 *)__LC_SUBCHANNEL_ID);
 941                 rc |= put_guest_lc(vcpu, inti->io.subchannel_nr,
 942                                 (u16 *)__LC_SUBCHANNEL_NR);
 943                 rc |= put_guest_lc(vcpu, inti->io.io_int_parm,
 944                                 (u32 *)__LC_IO_INT_PARM);
 945                 rc |= put_guest_lc(vcpu, inti->io.io_int_word,
 946                                 (u32 *)__LC_IO_INT_WORD);
 947                 rc |= write_guest_lc(vcpu, __LC_IO_OLD_PSW,
 948                                 &vcpu->arch.sie_block->gpsw,
 949                                 sizeof(psw_t));
 950                 rc |= read_guest_lc(vcpu, __LC_IO_NEW_PSW,
 951                                 &vcpu->arch.sie_block->gpsw,
 952                                 sizeof(psw_t));
 953                 kfree(inti);
 954         }
 955
 956         return rc ? -EFAULT : 0;
 957 }
 958
 959 typedef int (*deliver_irq_t)(struct kvm_vcpu *vcpu);
 960
 961 static const deliver_irq_t deliver_irq_funcs[] = {
 962         [IRQ_PEND_MCHK_EX]        = __deliver_machine_check,
 963         [IRQ_PEND_MCHK_REP]       = __deliver_machine_check,
 964         [IRQ_PEND_PROG]           = __deliver_prog,
 965         [IRQ_PEND_EXT_EMERGENCY]  = __deliver_emergency_signal,
 966         [IRQ_PEND_EXT_EXTERNAL]   = __deliver_external_call,
 967         [IRQ_PEND_EXT_CLOCK_COMP] = __deliver_ckc,
 968         [IRQ_PEND_EXT_CPU_TIMER]  = __deliver_cpu_timer,
 969         [IRQ_PEND_RESTART]        = __deliver_restart,
 970         [IRQ_PEND_SET_PREFIX]     = __deliver_set_prefix,
 971         [IRQ_PEND_PFAULT_INIT]    = __deliver_pfault_init,
 972         [IRQ_PEND_EXT_SERVICE]    = __deliver_service,
 973         [IRQ_PEND_PFAULT_DONE]    = __deliver_pfault_done,
 974         [IRQ_PEND_VIRTIO]         = __deliver_virtio,
 975 };
 976
 977 /* Check whether an external call is pending (deliverable or not) */
 978 int kvm_s390_ext_call_pending(struct kvm_vcpu *vcpu)
 979 {
 980         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
 981
 982         if (!sclp.has_sigpif)
 983                 return test_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs);
 984
 985         return sca_ext_call_pending(vcpu, NULL);
 986 }
 987
 988 int kvm_s390_vcpu_has_irq(struct kvm_vcpu *vcpu, int exclude_stop)
 989 {
 990         if (deliverable_irqs(vcpu))
 991                 return 1;
 992
 993         if (kvm_cpu_has_pending_timer(vcpu))
 994                 return 1;
 995
 996         /* external call pending and deliverable */
 997         if (kvm_s390_ext_call_pending(vcpu) &&
 998             !psw_extint_disabled(vcpu) &&
 999             (vcpu->arch.sie_block->gcr[0] & 0x2000ul))
1000                 return 1;
1001
1002         if (!exclude_stop && kvm_s390_is_stop_irq_pending(vcpu))
1003                 return 1;
1004         return 0;
1005 }
1006
1007 int kvm_cpu_has_pending_timer(struct kvm_vcpu *vcpu)
1008 {
1009         return ckc_irq_pending(vcpu) || cpu_timer_irq_pending(vcpu);
1010 }
1011
1012 static u64 __calculate_sltime(struct kvm_vcpu *vcpu)
1013 {
1014         const u64 now = kvm_s390_get_tod_clock_fast(vcpu->kvm);
1015         const u64 ckc = vcpu->arch.sie_block->ckc;
1016         u64 cputm, sltime = 0;
1017
1018         if (ckc_interrupts_enabled(vcpu)) {
1019                 if (vcpu->arch.sie_block->gcr[0] & 0x0020000000000000ul) {
1020                         if ((s64)now < (s64)ckc)
1021                                 sltime = tod_to_ns((s64)ckc - (s64)now);
1022                 } else if (now < ckc) {
1023                         sltime = tod_to_ns(ckc - now);
1024                 }
1025                 /* already expired */
1026                 if (!sltime)
1027                         return 0;
1028                 if (cpu_timer_interrupts_enabled(vcpu)) {
1029                         cputm = kvm_s390_get_cpu_timer(vcpu);
1030                         /* already expired? */
1031                         if (cputm >> 63)
1032                                 return 0;
1033                         return min(sltime, tod_to_ns(cputm));
1034                 }
1035         } else if (cpu_timer_interrupts_enabled(vcpu)) {
1036                 sltime = kvm_s390_get_cpu_timer(vcpu);
1037                 /* already expired? */
1038                 if (sltime >> 63)
1039                         return 0;
1040         }
1041         return sltime;
1042 }
1043
1044 int kvm_s390_handle_wait(struct kvm_vcpu *vcpu)
1045 {
1046         u64 sltime;
1047
1048         vcpu->stat.exit_wait_state++;
1049
1050         /* fast path */
1051         if (kvm_arch_vcpu_runnable(vcpu))
1052                 return 0;
1053
1054         if (psw_interrupts_disabled(vcpu)) {
1055                 VCPU_EVENT(vcpu, 3, "%s", "disabled wait");
1056                 return -EOPNOTSUPP; /* disabled wait */
1057         }
1058
1059         if (!ckc_interrupts_enabled(vcpu) &&
1060             !cpu_timer_interrupts_enabled(vcpu)) {
1061                 VCPU_EVENT(vcpu, 3, "%s", "enabled wait w/o timer");
1062                 __set_cpu_idle(vcpu);
1063                 goto no_timer;
1064         }
1065
1066         sltime = __calculate_sltime(vcpu);
1067         if (!sltime)
1068                 return 0;
1069
1070         __set_cpu_idle(vcpu);
1071         hrtimer_start(&vcpu->arch.ckc_timer, sltime, HRTIMER_MODE_REL);
1072         VCPU_EVENT(vcpu, 4, "enabled wait: %llu ns", sltime);
1073 no_timer:
1074         srcu_read_unlock(&vcpu->kvm->srcu, vcpu->srcu_idx);
1075         kvm_vcpu_block(vcpu);
1076         __unset_cpu_idle(vcpu);
1077         vcpu->srcu_idx = srcu_read_lock(&vcpu->kvm->srcu);
1078
1079         hrtimer_cancel(&vcpu->arch.ckc_timer);
1080         return 0;
1081 }
1082
1083 void kvm_s390_vcpu_wakeup(struct kvm_vcpu *vcpu)
1084 {
1085         /*
1086          * We cannot move this into the if, as the CPU might be already
1087          * in kvm_vcpu_block without having the waitqueue set (polling)
1088          */
1089         vcpu->valid_wakeup = true;
1090         if (swait_active(&vcpu->wq)) {
1091                 /*
1092                  * The vcpu gave up the cpu voluntarily, mark it as a good
1093                  * yield-candidate.
1094                  */
1095                 vcpu->preempted = true;
1096                 swake_up(&vcpu->wq);
1097                 vcpu->stat.halt_wakeup++;
1098         }
1099         /*
1100          * The VCPU might not be sleeping but is executing the VSIE. Let's
1101          * kick it, so it leaves the SIE to process the request.
1102          */
1103         kvm_s390_vsie_kick(vcpu);
1104 }
1105
1106 enum hrtimer_restart kvm_s390_idle_wakeup(struct hrtimer *timer)
1107 {
1108         struct kvm_vcpu *vcpu;
1109         u64 sltime;
1110
1111         vcpu = container_of(timer, struct kvm_vcpu, arch.ckc_timer);
1112         sltime = __calculate_sltime(vcpu);
1113
1114         /*
1115          * If the monotonic clock runs faster than the tod clock we might be
1116          * woken up too early and have to go back to sleep to avoid deadlocks.
1117          */
1118         if (sltime && hrtimer_forward_now(timer, ns_to_ktime(sltime)))
1119                 return HRTIMER_RESTART;
1120         kvm_s390_vcpu_wakeup(vcpu);
1121         return HRTIMER_NORESTART;
1122 }
1123
1124 void kvm_s390_clear_local_irqs(struct kvm_vcpu *vcpu)
1125 {
1126         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1127
1128         spin_lock(&li->lock);
1129         li->pending_irqs = 0;
1130         bitmap_zero(li->sigp_emerg_pending, KVM_MAX_VCPUS);
1131         memset(&li->irq, 0, sizeof(li->irq));
1132         spin_unlock(&li->lock);
1133
1134         sca_clear_ext_call(vcpu);
1135 }
1136
1137 int __must_check kvm_s390_deliver_pending_interrupts(struct kvm_vcpu *vcpu)
1138 {
1139         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1140         deliver_irq_t func;
1141         int rc = 0;
1142         unsigned long irq_type;
1143         unsigned long irqs;
1144
1145         __reset_intercept_indicators(vcpu);
1146
1147         /* pending ckc conditions might have been invalidated */
1148         clear_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1149         if (ckc_irq_pending(vcpu))
1150                 set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1151
1152         /* pending cpu timer conditions might have been invalidated */
1153         clear_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1154         if (cpu_timer_irq_pending(vcpu))
1155                 set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1156
1157         while ((irqs = deliverable_irqs(vcpu)) && !rc) {
1158                 /* bits are in the order of interrupt priority */
1159                 irq_type = find_first_bit(&irqs, IRQ_PEND_COUNT);
1160                 if (is_ioirq(irq_type)) {
1161                         rc = __deliver_io(vcpu, irq_type);
1162                 } else {
1163                         func = deliver_irq_funcs[irq_type];
1164                         if (!func) {
1165                                 WARN_ON_ONCE(func == NULL);
1166                                 clear_bit(irq_type, &li->pending_irqs);
1167                                 continue;
1168                         }
1169                         rc = func(vcpu);
1170                 }
1171         }
1172
1173         set_intercept_indicators(vcpu);
1174
1175         return rc;
1176 }
1177
1178 static int __inject_prog(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1179 {
1180         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1181
1182         VCPU_EVENT(vcpu, 3, "inject: program irq code 0x%x", irq->u.pgm.code);
1183         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_PROGRAM_INT,
1184                                    irq->u.pgm.code, 0);
1185
1186         if (!(irq->u.pgm.flags & KVM_S390_PGM_FLAGS_ILC_VALID)) {
1187                 /* auto detection if no valid ILC was given */
1188                 irq->u.pgm.flags &= ~KVM_S390_PGM_FLAGS_ILC_MASK;
1189                 irq->u.pgm.flags |= kvm_s390_get_ilen(vcpu);
1190                 irq->u.pgm.flags |= KVM_S390_PGM_FLAGS_ILC_VALID;
1191         }
1192
1193         if (irq->u.pgm.code == PGM_PER) {
1194                 li->irq.pgm.code |= PGM_PER;
1195                 li->irq.pgm.flags = irq->u.pgm.flags;
1196                 /* only modify PER related information */
1197                 li->irq.pgm.per_address = irq->u.pgm.per_address;
1198                 li->irq.pgm.per_code = irq->u.pgm.per_code;
1199                 li->irq.pgm.per_atmid = irq->u.pgm.per_atmid;
1200                 li->irq.pgm.per_access_id = irq->u.pgm.per_access_id;
1201         } else if (!(irq->u.pgm.code & PGM_PER)) {
1202                 li->irq.pgm.code = (li->irq.pgm.code & PGM_PER) |
1203                                    irq->u.pgm.code;
1204                 li->irq.pgm.flags = irq->u.pgm.flags;
1205                 /* only modify non-PER information */
1206                 li->irq.pgm.trans_exc_code = irq->u.pgm.trans_exc_code;
1207                 li->irq.pgm.mon_code = irq->u.pgm.mon_code;
1208                 li->irq.pgm.data_exc_code = irq->u.pgm.data_exc_code;
1209                 li->irq.pgm.mon_class_nr = irq->u.pgm.mon_class_nr;
1210                 li->irq.pgm.exc_access_id = irq->u.pgm.exc_access_id;
1211                 li->irq.pgm.op_access_id = irq->u.pgm.op_access_id;
1212         } else {
1213                 li->irq.pgm = irq->u.pgm;
1214         }
1215         set_bit(IRQ_PEND_PROG, &li->pending_irqs);
1216         return 0;
1217 }
1218
1219 static int __inject_pfault_init(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1220 {
1221         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1222
1223         VCPU_EVENT(vcpu, 4, "inject: pfault init parameter block at 0x%llx",
1224                    irq->u.ext.ext_params2);
1225         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_PFAULT_INIT,
1226                                    irq->u.ext.ext_params,
1227                                    irq->u.ext.ext_params2);
1228
1229         li->irq.ext = irq->u.ext;
1230         set_bit(IRQ_PEND_PFAULT_INIT, &li->pending_irqs);
1231         atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1232         return 0;
1233 }
1234
1235 static int __inject_extcall(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1236 {
1237         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1238         struct kvm_s390_extcall_info *extcall = &li->irq.extcall;
1239         uint16_t src_id = irq->u.extcall.code;
1240
1241         VCPU_EVENT(vcpu, 4, "inject: external call source-cpu:%u",
1242                    src_id);
1243         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EXTERNAL_CALL,
1244                                    src_id, 0);
1245
1246         /* sending vcpu invalid */
1247         if (kvm_get_vcpu_by_id(vcpu->kvm, src_id) == NULL)
1248                 return -EINVAL;
1249
1250         if (sclp.has_sigpif)
1251                 return sca_inject_ext_call(vcpu, src_id);
1252
1253         if (test_and_set_bit(IRQ_PEND_EXT_EXTERNAL, &li->pending_irqs))
1254                 return -EBUSY;
1255         *extcall = irq->u.extcall;
1256         atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1257         return 0;
1258 }
1259
1260 static int __inject_set_prefix(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1261 {
1262         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1263         struct kvm_s390_prefix_info *prefix = &li->irq.prefix;
1264
1265         VCPU_EVENT(vcpu, 3, "inject: set prefix to %x",
1266                    irq->u.prefix.address);
1267         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_SET_PREFIX,
1268                                    irq->u.prefix.address, 0);
1269
1270         if (!is_vcpu_stopped(vcpu))
1271                 return -EBUSY;
1272
1273         *prefix = irq->u.prefix;
1274         set_bit(IRQ_PEND_SET_PREFIX, &li->pending_irqs);
1275         return 0;
1276 }
1277
1278 #define KVM_S390_STOP_SUPP_FLAGS (KVM_S390_STOP_FLAG_STORE_STATUS)
1279 static int __inject_sigp_stop(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1280 {
1281         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1282         struct kvm_s390_stop_info *stop = &li->irq.stop;
1283         int rc = 0;
1284
1285         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_SIGP_STOP, 0, 0);
1286
1287         if (irq->u.stop.flags & ~KVM_S390_STOP_SUPP_FLAGS)
1288                 return -EINVAL;
1289
1290         if (is_vcpu_stopped(vcpu)) {
1291                 if (irq->u.stop.flags & KVM_S390_STOP_FLAG_STORE_STATUS)
1292                         rc = kvm_s390_store_status_unloaded(vcpu,
1293                                                 KVM_S390_STORE_STATUS_NOADDR);
1294                 return rc;
1295         }
1296
1297         if (test_and_set_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs))
1298                 return -EBUSY;
1299         stop->flags = irq->u.stop.flags;
1300         __set_cpuflag(vcpu, CPUSTAT_STOP_INT);
1301         return 0;
1302 }
1303
1304 static int __inject_sigp_restart(struct kvm_vcpu *vcpu,
1305                                  struct kvm_s390_irq *irq)
1306 {
1307         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1308
1309         VCPU_EVENT(vcpu, 3, "%s", "inject: restart int");
1310         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_RESTART, 0, 0);
1311
1312         set_bit(IRQ_PEND_RESTART, &li->pending_irqs);
1313         return 0;
1314 }
1315
1316 static int __inject_sigp_emergency(struct kvm_vcpu *vcpu,
1317                                    struct kvm_s390_irq *irq)
1318 {
1319         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1320
1321         VCPU_EVENT(vcpu, 4, "inject: emergency from cpu %u",
1322                    irq->u.emerg.code);
1323         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_EMERGENCY,
1324                                    irq->u.emerg.code, 0);
1325
1326         /* sending vcpu invalid */
1327         if (kvm_get_vcpu_by_id(vcpu->kvm, irq->u.emerg.code) == NULL)
1328                 return -EINVAL;
1329
1330         set_bit(irq->u.emerg.code, li->sigp_emerg_pending);
1331         set_bit(IRQ_PEND_EXT_EMERGENCY, &li->pending_irqs);
1332         atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1333         return 0;
1334 }
1335
1336 static int __inject_mchk(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1337 {
1338         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1339         struct kvm_s390_mchk_info *mchk = &li->irq.mchk;
1340
1341         VCPU_EVENT(vcpu, 3, "inject: machine check mcic 0x%llx",
1342                    irq->u.mchk.mcic);
1343         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_MCHK, 0,
1344                                    irq->u.mchk.mcic);
1345
1346         /*
1347          * Because repressible machine checks can be indicated along with
1348          * exigent machine checks (PoP, Chapter 11, Interruption action)
1349          * we need to combine cr14, mcic and external damage code.
1350          * Failing storage address and the logout area should not be or'ed
1351          * together, we just indicate the last occurrence of the corresponding
1352          * machine check
1353          */
1354         mchk->cr14 |= irq->u.mchk.cr14;
1355         mchk->mcic |= irq->u.mchk.mcic;
1356         mchk->ext_damage_code |= irq->u.mchk.ext_damage_code;
1357         mchk->failing_storage_address = irq->u.mchk.failing_storage_address;
1358         memcpy(&mchk->fixed_logout, &irq->u.mchk.fixed_logout,
1359                sizeof(mchk->fixed_logout));
1360         if (mchk->mcic & MCHK_EX_MASK)
1361                 set_bit(IRQ_PEND_MCHK_EX, &li->pending_irqs);
1362         else if (mchk->mcic & MCHK_REP_MASK)
1363                 set_bit(IRQ_PEND_MCHK_REP,  &li->pending_irqs);
1364         return 0;
1365 }
1366
1367 static int __inject_ckc(struct kvm_vcpu *vcpu)
1368 {
1369         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1370
1371         VCPU_EVENT(vcpu, 3, "%s", "inject: clock comparator external");
1372         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CLOCK_COMP,
1373                                    0, 0);
1374
1375         set_bit(IRQ_PEND_EXT_CLOCK_COMP, &li->pending_irqs);
1376         atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1377         return 0;
1378 }
1379
1380 static int __inject_cpu_timer(struct kvm_vcpu *vcpu)
1381 {
1382         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1383
1384         VCPU_EVENT(vcpu, 3, "%s", "inject: cpu timer external");
1385         trace_kvm_s390_inject_vcpu(vcpu->vcpu_id, KVM_S390_INT_CPU_TIMER,
1386                                    0, 0);
1387
1388         set_bit(IRQ_PEND_EXT_CPU_TIMER, &li->pending_irqs);
1389         atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1390         return 0;
1391 }
1392
1393 static struct kvm_s390_interrupt_info *get_io_int(struct kvm *kvm,
1394                                                   int isc, u32 schid)
1395 {
1396         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1397         struct list_head *isc_list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1398         struct kvm_s390_interrupt_info *iter;
1399         u16 id = (schid & 0xffff0000U) >> 16;
1400         u16 nr = schid & 0x0000ffffU;
1401
1402         spin_lock(&fi->lock);
1403         list_for_each_entry(iter, isc_list, list) {
1404                 if (schid && (id != iter->io.subchannel_id ||
1405                               nr != iter->io.subchannel_nr))
1406                         continue;
1407                 /* found an appropriate entry */
1408                 list_del_init(&iter->list);
1409                 fi->counters[FIRQ_CNTR_IO] -= 1;
1410                 if (list_empty(isc_list))
1411                         clear_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1412                 spin_unlock(&fi->lock);
1413                 return iter;
1414         }
1415         spin_unlock(&fi->lock);
1416         return NULL;
1417 }
1418
1419 /*
1420  * Dequeue and return an I/O interrupt matching any of the interruption
1421  * subclasses as designated by the isc mask in cr6 and the schid (if != 0).
1422  */
1423 struct kvm_s390_interrupt_info *kvm_s390_get_io_int(struct kvm *kvm,
1424                                                     u64 isc_mask, u32 schid)
1425 {
1426         struct kvm_s390_interrupt_info *inti = NULL;
1427         int isc;
1428
1429         for (isc = 0; isc <= MAX_ISC && !inti; isc++) {
1430                 if (isc_mask & isc_to_isc_bits(isc))
1431                         inti = get_io_int(kvm, isc, schid);
1432         }
1433         return inti;
1434 }
1435
1436 #define SCCB_MASK 0xFFFFFFF8
1437 #define SCCB_EVENT_PENDING 0x3
1438
1439 static int __inject_service(struct kvm *kvm,
1440                              struct kvm_s390_interrupt_info *inti)
1441 {
1442         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1443
1444         spin_lock(&fi->lock);
1445         fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_EVENT_PENDING;
1446         /*
1447          * Early versions of the QEMU s390 bios will inject several
1448          * service interrupts after another without handling a
1449          * condition code indicating busy.
1450          * We will silently ignore those superfluous sccb values.
1451          * A future version of QEMU will take care of serialization
1452          * of servc requests
1453          */
1454         if (fi->srv_signal.ext_params & SCCB_MASK)
1455                 goto out;
1456         fi->srv_signal.ext_params |= inti->ext.ext_params & SCCB_MASK;
1457         set_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs);
1458 out:
1459         spin_unlock(&fi->lock);
1460         kfree(inti);
1461         return 0;
1462 }
1463
1464 static int __inject_virtio(struct kvm *kvm,
1465                             struct kvm_s390_interrupt_info *inti)
1466 {
1467         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1468
1469         spin_lock(&fi->lock);
1470         if (fi->counters[FIRQ_CNTR_VIRTIO] >= KVM_S390_MAX_VIRTIO_IRQS) {
1471                 spin_unlock(&fi->lock);
1472                 return -EBUSY;
1473         }
1474         fi->counters[FIRQ_CNTR_VIRTIO] += 1;
1475         list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_VIRTIO]);
1476         set_bit(IRQ_PEND_VIRTIO, &fi->pending_irqs);
1477         spin_unlock(&fi->lock);
1478         return 0;
1479 }
1480
1481 static int __inject_pfault_done(struct kvm *kvm,
1482                                  struct kvm_s390_interrupt_info *inti)
1483 {
1484         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1485
1486         spin_lock(&fi->lock);
1487         if (fi->counters[FIRQ_CNTR_PFAULT] >=
1488                 (ASYNC_PF_PER_VCPU * KVM_MAX_VCPUS)) {
1489                 spin_unlock(&fi->lock);
1490                 return -EBUSY;
1491         }
1492         fi->counters[FIRQ_CNTR_PFAULT] += 1;
1493         list_add_tail(&inti->list, &fi->lists[FIRQ_LIST_PFAULT]);
1494         set_bit(IRQ_PEND_PFAULT_DONE, &fi->pending_irqs);
1495         spin_unlock(&fi->lock);
1496         return 0;
1497 }
1498
1499 #define CR_PENDING_SUBCLASS 28
1500 static int __inject_float_mchk(struct kvm *kvm,
1501                                 struct kvm_s390_interrupt_info *inti)
1502 {
1503         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1504
1505         spin_lock(&fi->lock);
1506         fi->mchk.cr14 |= inti->mchk.cr14 & (1UL << CR_PENDING_SUBCLASS);
1507         fi->mchk.mcic |= inti->mchk.mcic;
1508         set_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs);
1509         spin_unlock(&fi->lock);
1510         kfree(inti);
1511         return 0;
1512 }
1513
1514 static int __inject_io(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1515 {
1516         struct kvm_s390_float_interrupt *fi;
1517         struct list_head *list;
1518         int isc;
1519
1520         fi = &kvm->arch.float_int;
1521         spin_lock(&fi->lock);
1522         if (fi->counters[FIRQ_CNTR_IO] >= KVM_S390_MAX_FLOAT_IRQS) {
1523                 spin_unlock(&fi->lock);
1524                 return -EBUSY;
1525         }
1526         fi->counters[FIRQ_CNTR_IO] += 1;
1527
1528         if (inti->type & KVM_S390_INT_IO_AI_MASK)
1529                 VM_EVENT(kvm, 4, "%s", "inject: I/O (AI)");
1530         else
1531                 VM_EVENT(kvm, 4, "inject: I/O %x ss %x schid %04x",
1532                         inti->io.subchannel_id >> 8,
1533                         inti->io.subchannel_id >> 1 & 0x3,
1534                         inti->io.subchannel_nr);
1535         isc = int_word_to_isc(inti->io.io_int_word);
1536         list = &fi->lists[FIRQ_LIST_IO_ISC_0 + isc];
1537         list_add_tail(&inti->list, list);
1538         set_bit(IRQ_PEND_IO_ISC_0 + isc, &fi->pending_irqs);
1539         spin_unlock(&fi->lock);
1540         return 0;
1541 }
1542
1543 /*
1544  * Find a destination VCPU for a floating irq and kick it.
1545  */
1546 static void __floating_irq_kick(struct kvm *kvm, u64 type)
1547 {
1548         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1549         struct kvm_s390_local_interrupt *li;
1550         struct kvm_vcpu *dst_vcpu;
1551         int sigcpu, online_vcpus, nr_tries = 0;
1552
1553         online_vcpus = atomic_read(&kvm->online_vcpus);
1554         if (!online_vcpus)
1555                 return;
1556
1557         /* find idle VCPUs first, then round robin */
1558         sigcpu = find_first_bit(fi->idle_mask, online_vcpus);
1559         if (sigcpu == online_vcpus) {
1560                 do {
1561                         sigcpu = fi->next_rr_cpu;
1562                         fi->next_rr_cpu = (fi->next_rr_cpu + 1) % online_vcpus;
1563                         /* avoid endless loops if all vcpus are stopped */
1564                         if (nr_tries++ >= online_vcpus)
1565                                 return;
1566                 } while (is_vcpu_stopped(kvm_get_vcpu(kvm, sigcpu)));
1567         }
1568         dst_vcpu = kvm_get_vcpu(kvm, sigcpu);
1569
1570         /* make the VCPU drop out of the SIE, or wake it up if sleeping */
1571         li = &dst_vcpu->arch.local_int;
1572         spin_lock(&li->lock);
1573         switch (type) {
1574         case KVM_S390_MCHK:
1575                 atomic_or(CPUSTAT_STOP_INT, li->cpuflags);
1576                 break;
1577         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1578                 atomic_or(CPUSTAT_IO_INT, li->cpuflags);
1579                 break;
1580         default:
1581                 atomic_or(CPUSTAT_EXT_INT, li->cpuflags);
1582                 break;
1583         }
1584         spin_unlock(&li->lock);
1585         kvm_s390_vcpu_wakeup(dst_vcpu);
1586 }
1587
1588 static int __inject_vm(struct kvm *kvm, struct kvm_s390_interrupt_info *inti)
1589 {
1590         u64 type = READ_ONCE(inti->type);
1591         int rc;
1592
1593         switch (type) {
1594         case KVM_S390_MCHK:
1595                 rc = __inject_float_mchk(kvm, inti);
1596                 break;
1597         case KVM_S390_INT_VIRTIO:
1598                 rc = __inject_virtio(kvm, inti);
1599                 break;
1600         case KVM_S390_INT_SERVICE:
1601                 rc = __inject_service(kvm, inti);
1602                 break;
1603         case KVM_S390_INT_PFAULT_DONE:
1604                 rc = __inject_pfault_done(kvm, inti);
1605                 break;
1606         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1607                 rc = __inject_io(kvm, inti);
1608                 break;
1609         default:
1610                 rc = -EINVAL;
1611         }
1612         if (rc)
1613                 return rc;
1614
1615         __floating_irq_kick(kvm, type);
1616         return 0;
1617 }
1618
1619 int kvm_s390_inject_vm(struct kvm *kvm,
1620                        struct kvm_s390_interrupt *s390int)
1621 {
1622         struct kvm_s390_interrupt_info *inti;
1623         int rc;
1624
1625         inti = kzalloc(sizeof(*inti), GFP_KERNEL);
1626         if (!inti)
1627                 return -ENOMEM;
1628
1629         inti->type = s390int->type;
1630         switch (inti->type) {
1631         case KVM_S390_INT_VIRTIO:
1632                 VM_EVENT(kvm, 5, "inject: virtio parm:%x,parm64:%llx",
1633                          s390int->parm, s390int->parm64);
1634                 inti->ext.ext_params = s390int->parm;
1635                 inti->ext.ext_params2 = s390int->parm64;
1636                 break;
1637         case KVM_S390_INT_SERVICE:
1638                 VM_EVENT(kvm, 4, "inject: sclp parm:%x", s390int->parm);
1639                 inti->ext.ext_params = s390int->parm;
1640                 break;
1641         case KVM_S390_INT_PFAULT_DONE:
1642                 inti->ext.ext_params2 = s390int->parm64;
1643                 break;
1644         case KVM_S390_MCHK:
1645                 VM_EVENT(kvm, 3, "inject: machine check mcic 0x%llx",
1646                          s390int->parm64);
1647                 inti->mchk.cr14 = s390int->parm; /* upper bits are not used */
1648                 inti->mchk.mcic = s390int->parm64;
1649                 break;
1650         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1651                 inti->io.subchannel_id = s390int->parm >> 16;
1652                 inti->io.subchannel_nr = s390int->parm & 0x0000ffffu;
1653                 inti->io.io_int_parm = s390int->parm64 >> 32;
1654                 inti->io.io_int_word = s390int->parm64 & 0x00000000ffffffffull;
1655                 break;
1656         default:
1657                 kfree(inti);
1658                 return -EINVAL;
1659         }
1660         trace_kvm_s390_inject_vm(s390int->type, s390int->parm, s390int->parm64,
1661                                  2);
1662
1663         rc = __inject_vm(kvm, inti);
1664         if (rc)
1665                 kfree(inti);
1666         return rc;
1667 }
1668
1669 int kvm_s390_reinject_io_int(struct kvm *kvm,
1670                               struct kvm_s390_interrupt_info *inti)
1671 {
1672         return __inject_vm(kvm, inti);
1673 }
1674
1675 int s390int_to_s390irq(struct kvm_s390_interrupt *s390int,
1676                        struct kvm_s390_irq *irq)
1677 {
1678         irq->type = s390int->type;
1679         switch (irq->type) {
1680         case KVM_S390_PROGRAM_INT:
1681                 if (s390int->parm & 0xffff0000)
1682                         return -EINVAL;
1683                 irq->u.pgm.code = s390int->parm;
1684                 break;
1685         case KVM_S390_SIGP_SET_PREFIX:
1686                 irq->u.prefix.address = s390int->parm;
1687                 break;
1688         case KVM_S390_SIGP_STOP:
1689                 irq->u.stop.flags = s390int->parm;
1690                 break;
1691         case KVM_S390_INT_EXTERNAL_CALL:
1692                 if (s390int->parm & 0xffff0000)
1693                         return -EINVAL;
1694                 irq->u.extcall.code = s390int->parm;
1695                 break;
1696         case KVM_S390_INT_EMERGENCY:
1697                 if (s390int->parm & 0xffff0000)
1698                         return -EINVAL;
1699                 irq->u.emerg.code = s390int->parm;
1700                 break;
1701         case KVM_S390_MCHK:
1702                 irq->u.mchk.mcic = s390int->parm64;
1703                 break;
1704         case KVM_S390_INT_PFAULT_INIT:
1705                 irq->u.ext.ext_params = s390int->parm;
1706                 irq->u.ext.ext_params2 = s390int->parm64;
1707                 break;
1708         case KVM_S390_RESTART:
1709         case KVM_S390_INT_CLOCK_COMP:
1710         case KVM_S390_INT_CPU_TIMER:
1711                 break;
1712         default:
1713                 return -EINVAL;
1714         }
1715         return 0;
1716 }
1717
1718 int kvm_s390_is_stop_irq_pending(struct kvm_vcpu *vcpu)
1719 {
1720         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1721
1722         return test_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1723 }
1724
1725 void kvm_s390_clear_stop_irq(struct kvm_vcpu *vcpu)
1726 {
1727         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1728
1729         spin_lock(&li->lock);
1730         li->irq.stop.flags = 0;
1731         clear_bit(IRQ_PEND_SIGP_STOP, &li->pending_irqs);
1732         spin_unlock(&li->lock);
1733 }
1734
1735 static int do_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1736 {
1737         int rc;
1738
1739         switch (irq->type) {
1740         case KVM_S390_PROGRAM_INT:
1741                 rc = __inject_prog(vcpu, irq);
1742                 break;
1743         case KVM_S390_SIGP_SET_PREFIX:
1744                 rc = __inject_set_prefix(vcpu, irq);
1745                 break;
1746         case KVM_S390_SIGP_STOP:
1747                 rc = __inject_sigp_stop(vcpu, irq);
1748                 break;
1749         case KVM_S390_RESTART:
1750                 rc = __inject_sigp_restart(vcpu, irq);
1751                 break;
1752         case KVM_S390_INT_CLOCK_COMP:
1753                 rc = __inject_ckc(vcpu);
1754                 break;
1755         case KVM_S390_INT_CPU_TIMER:
1756                 rc = __inject_cpu_timer(vcpu);
1757                 break;
1758         case KVM_S390_INT_EXTERNAL_CALL:
1759                 rc = __inject_extcall(vcpu, irq);
1760                 break;
1761         case KVM_S390_INT_EMERGENCY:
1762                 rc = __inject_sigp_emergency(vcpu, irq);
1763                 break;
1764         case KVM_S390_MCHK:
1765                 rc = __inject_mchk(vcpu, irq);
1766                 break;
1767         case KVM_S390_INT_PFAULT_INIT:
1768                 rc = __inject_pfault_init(vcpu, irq);
1769                 break;
1770         case KVM_S390_INT_VIRTIO:
1771         case KVM_S390_INT_SERVICE:
1772         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1773         default:
1774                 rc = -EINVAL;
1775         }
1776
1777         return rc;
1778 }
1779
1780 int kvm_s390_inject_vcpu(struct kvm_vcpu *vcpu, struct kvm_s390_irq *irq)
1781 {
1782         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
1783         int rc;
1784
1785         spin_lock(&li->lock);
1786         rc = do_inject_vcpu(vcpu, irq);
1787         spin_unlock(&li->lock);
1788         if (!rc)
1789                 kvm_s390_vcpu_wakeup(vcpu);
1790         return rc;
1791 }
1792
1793 static inline void clear_irq_list(struct list_head *_list)
1794 {
1795         struct kvm_s390_interrupt_info *inti, *n;
1796
1797         list_for_each_entry_safe(inti, n, _list, list) {
1798                 list_del(&inti->list);
1799                 kfree(inti);
1800         }
1801 }
1802
1803 static void inti_to_irq(struct kvm_s390_interrupt_info *inti,
1804                        struct kvm_s390_irq *irq)
1805 {
1806         irq->type = inti->type;
1807         switch (inti->type) {
1808         case KVM_S390_INT_PFAULT_INIT:
1809         case KVM_S390_INT_PFAULT_DONE:
1810         case KVM_S390_INT_VIRTIO:
1811                 irq->u.ext = inti->ext;
1812                 break;
1813         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1814                 irq->u.io = inti->io;
1815                 break;
1816         }
1817 }
1818
1819 void kvm_s390_clear_float_irqs(struct kvm *kvm)
1820 {
1821         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1822         int i;
1823
1824         spin_lock(&fi->lock);
1825         fi->pending_irqs = 0;
1826         memset(&fi->srv_signal, 0, sizeof(fi->srv_signal));
1827         memset(&fi->mchk, 0, sizeof(fi->mchk));
1828         for (i = 0; i < FIRQ_LIST_COUNT; i++)
1829                 clear_irq_list(&fi->lists[i]);
1830         for (i = 0; i < FIRQ_MAX_COUNT; i++)
1831                 fi->counters[i] = 0;
1832         spin_unlock(&fi->lock);
1833 };
1834
1835 static int get_all_floating_irqs(struct kvm *kvm, u8 __user *usrbuf, u64 len)
1836 {
1837         struct kvm_s390_interrupt_info *inti;
1838         struct kvm_s390_float_interrupt *fi;
1839         struct kvm_s390_irq *buf;
1840         struct kvm_s390_irq *irq;
1841         int max_irqs;
1842         int ret = 0;
1843         int n = 0;
1844         int i;
1845
1846         if (len > KVM_S390_FLIC_MAX_BUFFER || len == 0)
1847                 return -EINVAL;
1848
1849         /*
1850          * We are already using -ENOMEM to signal
1851          * userspace it may retry with a bigger buffer,
1852          * so we need to use something else for this case
1853          */
1854         buf = vzalloc(len);
1855         if (!buf)
1856                 return -ENOBUFS;
1857
1858         max_irqs = len / sizeof(struct kvm_s390_irq);
1859
1860         fi = &kvm->arch.float_int;
1861         spin_lock(&fi->lock);
1862         for (i = 0; i < FIRQ_LIST_COUNT; i++) {
1863                 list_for_each_entry(inti, &fi->lists[i], list) {
1864                         if (n == max_irqs) {
1865                                 /* signal userspace to try again */
1866                                 ret = -ENOMEM;
1867                                 goto out;
1868                         }
1869                         inti_to_irq(inti, &buf[n]);
1870                         n++;
1871                 }
1872         }
1873         if (test_bit(IRQ_PEND_EXT_SERVICE, &fi->pending_irqs)) {
1874                 if (n == max_irqs) {
1875                         /* signal userspace to try again */
1876                         ret = -ENOMEM;
1877                         goto out;
1878                 }
1879                 irq = (struct kvm_s390_irq *) &buf[n];
1880                 irq->type = KVM_S390_INT_SERVICE;
1881                 irq->u.ext = fi->srv_signal;
1882                 n++;
1883         }
1884         if (test_bit(IRQ_PEND_MCHK_REP, &fi->pending_irqs)) {
1885                 if (n == max_irqs) {
1886                                 /* signal userspace to try again */
1887                                 ret = -ENOMEM;
1888                                 goto out;
1889                 }
1890                 irq = (struct kvm_s390_irq *) &buf[n];
1891                 irq->type = KVM_S390_MCHK;
1892                 irq->u.mchk = fi->mchk;
1893                 n++;
1894 }
1895
1896 out:
1897         spin_unlock(&fi->lock);
1898         if (!ret && n > 0) {
1899                 if (copy_to_user(usrbuf, buf, sizeof(struct kvm_s390_irq) * n))
1900                         ret = -EFAULT;
1901         }
1902         vfree(buf);
1903
1904         return ret < 0 ? ret : n;
1905 }
1906
1907 static int flic_ais_mode_get_all(struct kvm *kvm, struct kvm_device_attr *attr)
1908 {
1909         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
1910         struct kvm_s390_ais_all ais;
1911
1912         if (attr->attr < sizeof(ais))
1913                 return -EINVAL;
1914
1915         if (!test_kvm_facility(kvm, 72))
1916                 return -EOPNOTSUPP;
1917
1918         mutex_lock(&fi->ais_lock);
1919         ais.simm = fi->simm;
1920         ais.nimm = fi->nimm;
1921         mutex_unlock(&fi->ais_lock);
1922
1923         if (copy_to_user((void __user *)attr->addr, &ais, sizeof(ais)))
1924                 return -EFAULT;
1925
1926         return 0;
1927 }
1928
1929 static int flic_get_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
1930 {
1931         int r;
1932
1933         switch (attr->group) {
1934         case KVM_DEV_FLIC_GET_ALL_IRQS:
1935                 r = get_all_floating_irqs(dev->kvm, (u8 __user *) attr->addr,
1936                                           attr->attr);
1937                 break;
1938         case KVM_DEV_FLIC_AISM_ALL:
1939                 r = flic_ais_mode_get_all(dev->kvm, attr);
1940                 break;
1941         default:
1942                 r = -EINVAL;
1943         }
1944
1945         return r;
1946 }
1947
1948 static inline int copy_irq_from_user(struct kvm_s390_interrupt_info *inti,
1949                                      u64 addr)
1950 {
1951         struct kvm_s390_irq __user *uptr = (struct kvm_s390_irq __user *) addr;
1952         void *target = NULL;
1953         void __user *source;
1954         u64 size;
1955
1956         if (get_user(inti->type, (u64 __user *)addr))
1957                 return -EFAULT;
1958
1959         switch (inti->type) {
1960         case KVM_S390_INT_PFAULT_INIT:
1961         case KVM_S390_INT_PFAULT_DONE:
1962         case KVM_S390_INT_VIRTIO:
1963         case KVM_S390_INT_SERVICE:
1964                 target = (void *) &inti->ext;
1965                 source = &uptr->u.ext;
1966                 size = sizeof(inti->ext);
1967                 break;
1968         case KVM_S390_INT_IO_MIN...KVM_S390_INT_IO_MAX:
1969                 target = (void *) &inti->io;
1970                 source = &uptr->u.io;
1971                 size = sizeof(inti->io);
1972                 break;
1973         case KVM_S390_MCHK:
1974                 target = (void *) &inti->mchk;
1975                 source = &uptr->u.mchk;
1976                 size = sizeof(inti->mchk);
1977                 break;
1978         default:
1979                 return -EINVAL;
1980         }
1981
1982         if (copy_from_user(target, source, size))
1983                 return -EFAULT;
1984
1985         return 0;
1986 }
1987
1988 static int enqueue_floating_irq(struct kvm_device *dev,
1989                                 struct kvm_device_attr *attr)
1990 {
1991         struct kvm_s390_interrupt_info *inti = NULL;
1992         int r = 0;
1993         int len = attr->attr;
1994
1995         if (len % sizeof(struct kvm_s390_irq) != 0)
1996                 return -EINVAL;
1997         else if (len > KVM_S390_FLIC_MAX_BUFFER)
1998                 return -EINVAL;
1999
2000         while (len >= sizeof(struct kvm_s390_irq)) {
2001                 inti = kzalloc(sizeof(*inti), GFP_KERNEL);
2002                 if (!inti)
2003                         return -ENOMEM;
2004
2005                 r = copy_irq_from_user(inti, attr->addr);
2006                 if (r) {
2007                         kfree(inti);
2008                         return r;
2009                 }
2010                 r = __inject_vm(dev->kvm, inti);
2011                 if (r) {
2012                         kfree(inti);
2013                         return r;
2014                 }
2015                 len -= sizeof(struct kvm_s390_irq);
2016                 attr->addr += sizeof(struct kvm_s390_irq);
2017         }
2018
2019         return r;
2020 }
2021
2022 static struct s390_io_adapter *get_io_adapter(struct kvm *kvm, unsigned int id)
2023 {
2024         if (id >= MAX_S390_IO_ADAPTERS)
2025                 return NULL;
2026         return kvm->arch.adapters[id];
2027 }
2028
2029 static int register_io_adapter(struct kvm_device *dev,
2030                                struct kvm_device_attr *attr)
2031 {
2032         struct s390_io_adapter *adapter;
2033         struct kvm_s390_io_adapter adapter_info;
2034
2035         if (copy_from_user(&adapter_info,
2036                            (void __user *)attr->addr, sizeof(adapter_info)))
2037                 return -EFAULT;
2038
2039         if ((adapter_info.id >= MAX_S390_IO_ADAPTERS) ||
2040             (dev->kvm->arch.adapters[adapter_info.id] != NULL))
2041                 return -EINVAL;
2042
2043         adapter = kzalloc(sizeof(*adapter), GFP_KERNEL);
2044         if (!adapter)
2045                 return -ENOMEM;
2046
2047         INIT_LIST_HEAD(&adapter->maps);
2048         init_rwsem(&adapter->maps_lock);
2049         atomic_set(&adapter->nr_maps, 0);
2050         adapter->id = adapter_info.id;
2051         adapter->isc = adapter_info.isc;
2052         adapter->maskable = adapter_info.maskable;
2053         adapter->masked = false;
2054         adapter->swap = adapter_info.swap;
2055         adapter->suppressible = (adapter_info.flags) &
2056                                 KVM_S390_ADAPTER_SUPPRESSIBLE;
2057         dev->kvm->arch.adapters[adapter->id] = adapter;
2058
2059         return 0;
2060 }
2061
2062 int kvm_s390_mask_adapter(struct kvm *kvm, unsigned int id, bool masked)
2063 {
2064         int ret;
2065         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2066
2067         if (!adapter || !adapter->maskable)
2068                 return -EINVAL;
2069         ret = adapter->masked;
2070         adapter->masked = masked;
2071         return ret;
2072 }
2073
2074 static int kvm_s390_adapter_map(struct kvm *kvm, unsigned int id, __u64 addr)
2075 {
2076         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2077         struct s390_map_info *map;
2078         int ret;
2079
2080         if (!adapter || !addr)
2081                 return -EINVAL;
2082
2083         map = kzalloc(sizeof(*map), GFP_KERNEL);
2084         if (!map) {
2085                 ret = -ENOMEM;
2086                 goto out;
2087         }
2088         INIT_LIST_HEAD(&map->list);
2089         map->guest_addr = addr;
2090         map->addr = gmap_translate(kvm->arch.gmap, addr);
2091         if (map->addr == -EFAULT) {
2092                 ret = -EFAULT;
2093                 goto out;
2094         }
2095         ret = get_user_pages_fast(map->addr, 1, 1, &map->page);
2096         if (ret < 0)
2097                 goto out;
2098         BUG_ON(ret != 1);
2099         down_write(&adapter->maps_lock);
2100         if (atomic_inc_return(&adapter->nr_maps) < MAX_S390_ADAPTER_MAPS) {
2101                 list_add_tail(&map->list, &adapter->maps);
2102                 ret = 0;
2103         } else {
2104                 put_page(map->page);
2105                 ret = -EINVAL;
2106         }
2107         up_write(&adapter->maps_lock);
2108 out:
2109         if (ret)
2110                 kfree(map);
2111         return ret;
2112 }
2113
2114 static int kvm_s390_adapter_unmap(struct kvm *kvm, unsigned int id, __u64 addr)
2115 {
2116         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2117         struct s390_map_info *map, *tmp;
2118         int found = 0;
2119
2120         if (!adapter || !addr)
2121                 return -EINVAL;
2122
2123         down_write(&adapter->maps_lock);
2124         list_for_each_entry_safe(map, tmp, &adapter->maps, list) {
2125                 if (map->guest_addr == addr) {
2126                         found = 1;
2127                         atomic_dec(&adapter->nr_maps);
2128                         list_del(&map->list);
2129                         put_page(map->page);
2130                         kfree(map);
2131                         break;
2132                 }
2133         }
2134         up_write(&adapter->maps_lock);
2135
2136         return found ? 0 : -EINVAL;
2137 }
2138
2139 void kvm_s390_destroy_adapters(struct kvm *kvm)
2140 {
2141         int i;
2142         struct s390_map_info *map, *tmp;
2143
2144         for (i = 0; i < MAX_S390_IO_ADAPTERS; i++) {
2145                 if (!kvm->arch.adapters[i])
2146                         continue;
2147                 list_for_each_entry_safe(map, tmp,
2148                                          &kvm->arch.adapters[i]->maps, list) {
2149                         list_del(&map->list);
2150                         put_page(map->page);
2151                         kfree(map);
2152                 }
2153                 kfree(kvm->arch.adapters[i]);
2154         }
2155 }
2156
2157 static int modify_io_adapter(struct kvm_device *dev,
2158                              struct kvm_device_attr *attr)
2159 {
2160         struct kvm_s390_io_adapter_req req;
2161         struct s390_io_adapter *adapter;
2162         int ret;
2163
2164         if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2165                 return -EFAULT;
2166
2167         adapter = get_io_adapter(dev->kvm, req.id);
2168         if (!adapter)
2169                 return -EINVAL;
2170         switch (req.type) {
2171         case KVM_S390_IO_ADAPTER_MASK:
2172                 ret = kvm_s390_mask_adapter(dev->kvm, req.id, req.mask);
2173                 if (ret > 0)
2174                         ret = 0;
2175                 break;
2176         case KVM_S390_IO_ADAPTER_MAP:
2177                 ret = kvm_s390_adapter_map(dev->kvm, req.id, req.addr);
2178                 break;
2179         case KVM_S390_IO_ADAPTER_UNMAP:
2180                 ret = kvm_s390_adapter_unmap(dev->kvm, req.id, req.addr);
2181                 break;
2182         default:
2183                 ret = -EINVAL;
2184         }
2185
2186         return ret;
2187 }
2188
2189 static int clear_io_irq(struct kvm *kvm, struct kvm_device_attr *attr)
2190
2191 {
2192         const u64 isc_mask = 0xffUL << 24; /* all iscs set */
2193         u32 schid;
2194
2195         if (attr->flags)
2196                 return -EINVAL;
2197         if (attr->attr != sizeof(schid))
2198                 return -EINVAL;
2199         if (copy_from_user(&schid, (void __user *) attr->addr, sizeof(schid)))
2200                 return -EFAULT;
2201         kfree(kvm_s390_get_io_int(kvm, isc_mask, schid));
2202         /*
2203          * If userspace is conforming to the architecture, we can have at most
2204          * one pending I/O interrupt per subchannel, so this is effectively a
2205          * clear all.
2206          */
2207         return 0;
2208 }
2209
2210 static int modify_ais_mode(struct kvm *kvm, struct kvm_device_attr *attr)
2211 {
2212         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2213         struct kvm_s390_ais_req req;
2214         int ret = 0;
2215
2216         if (!test_kvm_facility(kvm, 72))
2217                 return -EOPNOTSUPP;
2218
2219         if (copy_from_user(&req, (void __user *)attr->addr, sizeof(req)))
2220                 return -EFAULT;
2221
2222         if (req.isc > MAX_ISC)
2223                 return -EINVAL;
2224
2225         trace_kvm_s390_modify_ais_mode(req.isc,
2226                                        (fi->simm & AIS_MODE_MASK(req.isc)) ?
2227                                        (fi->nimm & AIS_MODE_MASK(req.isc)) ?
2228                                        2 : KVM_S390_AIS_MODE_SINGLE :
2229                                        KVM_S390_AIS_MODE_ALL, req.mode);
2230
2231         mutex_lock(&fi->ais_lock);
2232         switch (req.mode) {
2233         case KVM_S390_AIS_MODE_ALL:
2234                 fi->simm &= ~AIS_MODE_MASK(req.isc);
2235                 fi->nimm &= ~AIS_MODE_MASK(req.isc);
2236                 break;
2237         case KVM_S390_AIS_MODE_SINGLE:
2238                 fi->simm |= AIS_MODE_MASK(req.isc);
2239                 fi->nimm &= ~AIS_MODE_MASK(req.isc);
2240                 break;
2241         default:
2242                 ret = -EINVAL;
2243         }
2244         mutex_unlock(&fi->ais_lock);
2245
2246         return ret;
2247 }
2248
2249 static int kvm_s390_inject_airq(struct kvm *kvm,
2250                                 struct s390_io_adapter *adapter)
2251 {
2252         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2253         struct kvm_s390_interrupt s390int = {
2254                 .type = KVM_S390_INT_IO(1, 0, 0, 0),
2255                 .parm = 0,
2256                 .parm64 = (adapter->isc << 27) | 0x80000000,
2257         };
2258         int ret = 0;
2259
2260         if (!test_kvm_facility(kvm, 72) || !adapter->suppressible)
2261                 return kvm_s390_inject_vm(kvm, &s390int);
2262
2263         mutex_lock(&fi->ais_lock);
2264         if (fi->nimm & AIS_MODE_MASK(adapter->isc)) {
2265                 trace_kvm_s390_airq_suppressed(adapter->id, adapter->isc);
2266                 goto out;
2267         }
2268
2269         ret = kvm_s390_inject_vm(kvm, &s390int);
2270         if (!ret && (fi->simm & AIS_MODE_MASK(adapter->isc))) {
2271                 fi->nimm |= AIS_MODE_MASK(adapter->isc);
2272                 trace_kvm_s390_modify_ais_mode(adapter->isc,
2273                                                KVM_S390_AIS_MODE_SINGLE, 2);
2274         }
2275 out:
2276         mutex_unlock(&fi->ais_lock);
2277         return ret;
2278 }
2279
2280 static int flic_inject_airq(struct kvm *kvm, struct kvm_device_attr *attr)
2281 {
2282         unsigned int id = attr->attr;
2283         struct s390_io_adapter *adapter = get_io_adapter(kvm, id);
2284
2285         if (!adapter)
2286                 return -EINVAL;
2287
2288         return kvm_s390_inject_airq(kvm, adapter);
2289 }
2290
2291 static int flic_ais_mode_set_all(struct kvm *kvm, struct kvm_device_attr *attr)
2292 {
2293         struct kvm_s390_float_interrupt *fi = &kvm->arch.float_int;
2294         struct kvm_s390_ais_all ais;
2295
2296         if (!test_kvm_facility(kvm, 72))
2297                 return -EOPNOTSUPP;
2298
2299         if (copy_from_user(&ais, (void __user *)attr->addr, sizeof(ais)))
2300                 return -EFAULT;
2301
2302         mutex_lock(&fi->ais_lock);
2303         fi->simm = ais.simm;
2304         fi->nimm = ais.nimm;
2305         mutex_unlock(&fi->ais_lock);
2306
2307         return 0;
2308 }
2309
2310 static int flic_set_attr(struct kvm_device *dev, struct kvm_device_attr *attr)
2311 {
2312         int r = 0;
2313         unsigned int i;
2314         struct kvm_vcpu *vcpu;
2315
2316         switch (attr->group) {
2317         case KVM_DEV_FLIC_ENQUEUE:
2318                 r = enqueue_floating_irq(dev, attr);
2319                 break;
2320         case KVM_DEV_FLIC_CLEAR_IRQS:
2321                 kvm_s390_clear_float_irqs(dev->kvm);
2322                 break;
2323         case KVM_DEV_FLIC_APF_ENABLE:
2324                 dev->kvm->arch.gmap->pfault_enabled = 1;
2325                 break;
2326         case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2327                 dev->kvm->arch.gmap->pfault_enabled = 0;
2328                 /*
2329                  * Make sure no async faults are in transition when
2330                  * clearing the queues. So we don't need to worry
2331                  * about late coming workers.
2332                  */
2333                 synchronize_srcu(&dev->kvm->srcu);
2334                 kvm_for_each_vcpu(i, vcpu, dev->kvm)
2335                         kvm_clear_async_pf_completion_queue(vcpu);
2336                 break;
2337         case KVM_DEV_FLIC_ADAPTER_REGISTER:
2338                 r = register_io_adapter(dev, attr);
2339                 break;
2340         case KVM_DEV_FLIC_ADAPTER_MODIFY:
2341                 r = modify_io_adapter(dev, attr);
2342                 break;
2343         case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2344                 r = clear_io_irq(dev->kvm, attr);
2345                 break;
2346         case KVM_DEV_FLIC_AISM:
2347                 r = modify_ais_mode(dev->kvm, attr);
2348                 break;
2349         case KVM_DEV_FLIC_AIRQ_INJECT:
2350                 r = flic_inject_airq(dev->kvm, attr);
2351                 break;
2352         case KVM_DEV_FLIC_AISM_ALL:
2353                 r = flic_ais_mode_set_all(dev->kvm, attr);
2354                 break;
2355         default:
2356                 r = -EINVAL;
2357         }
2358
2359         return r;
2360 }
2361
2362 static int flic_has_attr(struct kvm_device *dev,
2363                              struct kvm_device_attr *attr)
2364 {
2365         switch (attr->group) {
2366         case KVM_DEV_FLIC_GET_ALL_IRQS:
2367         case KVM_DEV_FLIC_ENQUEUE:
2368         case KVM_DEV_FLIC_CLEAR_IRQS:
2369         case KVM_DEV_FLIC_APF_ENABLE:
2370         case KVM_DEV_FLIC_APF_DISABLE_WAIT:
2371         case KVM_DEV_FLIC_ADAPTER_REGISTER:
2372         case KVM_DEV_FLIC_ADAPTER_MODIFY:
2373         case KVM_DEV_FLIC_CLEAR_IO_IRQ:
2374         case KVM_DEV_FLIC_AISM:
2375         case KVM_DEV_FLIC_AIRQ_INJECT:
2376         case KVM_DEV_FLIC_AISM_ALL:
2377                 return 0;
2378         }
2379         return -ENXIO;
2380 }
2381
2382 static int flic_create(struct kvm_device *dev, u32 type)
2383 {
2384         if (!dev)
2385                 return -EINVAL;
2386         if (dev->kvm->arch.flic)
2387                 return -EINVAL;
2388         dev->kvm->arch.flic = dev;
2389         return 0;
2390 }
2391
2392 static void flic_destroy(struct kvm_device *dev)
2393 {
2394         dev->kvm->arch.flic = NULL;
2395         kfree(dev);
2396 }
2397
2398 /* s390 floating irq controller (flic) */
2399 struct kvm_device_ops kvm_flic_ops = {
2400         .name = "kvm-flic",
2401         .get_attr = flic_get_attr,
2402         .set_attr = flic_set_attr,
2403         .has_attr = flic_has_attr,
2404         .create = flic_create,
2405         .destroy = flic_destroy,
2406 };
2407
2408 static unsigned long get_ind_bit(__u64 addr, unsigned long bit_nr, bool swap)
2409 {
2410         unsigned long bit;
2411
2412         bit = bit_nr + (addr % PAGE_SIZE) * 8;
2413
2414         return swap ? (bit ^ (BITS_PER_LONG - 1)) : bit;
2415 }
2416
2417 static struct s390_map_info *get_map_info(struct s390_io_adapter *adapter,
2418                                           u64 addr)
2419 {
2420         struct s390_map_info *map;
2421
2422         if (!adapter)
2423                 return NULL;
2424
2425         list_for_each_entry(map, &adapter->maps, list) {
2426                 if (map->guest_addr == addr)
2427                         return map;
2428         }
2429         return NULL;
2430 }
2431
2432 static int adapter_indicators_set(struct kvm *kvm,
2433                                   struct s390_io_adapter *adapter,
2434                                   struct kvm_s390_adapter_int *adapter_int)
2435 {
2436         unsigned long bit;
2437         int summary_set, idx;
2438         struct s390_map_info *info;
2439         void *map;
2440
2441         info = get_map_info(adapter, adapter_int->ind_addr);
2442         if (!info)
2443                 return -1;
2444         map = page_address(info->page);
2445         bit = get_ind_bit(info->addr, adapter_int->ind_offset, adapter->swap);
2446         set_bit(bit, map);
2447         idx = srcu_read_lock(&kvm->srcu);
2448         mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2449         set_page_dirty_lock(info->page);
2450         info = get_map_info(adapter, adapter_int->summary_addr);
2451         if (!info) {
2452                 srcu_read_unlock(&kvm->srcu, idx);
2453                 return -1;
2454         }
2455         map = page_address(info->page);
2456         bit = get_ind_bit(info->addr, adapter_int->summary_offset,
2457                           adapter->swap);
2458         summary_set = test_and_set_bit(bit, map);
2459         mark_page_dirty(kvm, info->guest_addr >> PAGE_SHIFT);
2460         set_page_dirty_lock(info->page);
2461         srcu_read_unlock(&kvm->srcu, idx);
2462         return summary_set ? 0 : 1;
2463 }
2464
2465 /*
2466  * < 0 - not injected due to error
2467  * = 0 - coalesced, summary indicator already active
2468  * > 0 - injected interrupt
2469  */
2470 static int set_adapter_int(struct kvm_kernel_irq_routing_entry *e,
2471                            struct kvm *kvm, int irq_source_id, int level,
2472                            bool line_status)
2473 {
2474         int ret;
2475         struct s390_io_adapter *adapter;
2476
2477         /* We're only interested in the 0->1 transition. */
2478         if (!level)
2479                 return 0;
2480         adapter = get_io_adapter(kvm, e->adapter.adapter_id);
2481         if (!adapter)
2482                 return -1;
2483         down_read(&adapter->maps_lock);
2484         ret = adapter_indicators_set(kvm, adapter, &e->adapter);
2485         up_read(&adapter->maps_lock);
2486         if ((ret > 0) && !adapter->masked) {
2487                 ret = kvm_s390_inject_airq(kvm, adapter);
2488                 if (ret == 0)
2489                         ret = 1;
2490         }
2491         return ret;
2492 }
2493
2494 /*
2495  * Inject the machine check to the guest.
2496  */
2497 void kvm_s390_reinject_machine_check(struct kvm_vcpu *vcpu,
2498                                      struct mcck_volatile_info *mcck_info)
2499 {
2500         struct kvm_s390_interrupt_info inti;
2501         struct kvm_s390_irq irq;
2502         struct kvm_s390_mchk_info *mchk;
2503         union mci mci;
2504         __u64 cr14 = 0;         /* upper bits are not used */
2505         int rc;
2506
2507         mci.val = mcck_info->mcic;
2508         if (mci.sr)
2509                 cr14 |= MCCK_CR14_RECOVERY_SUB_MASK;
2510         if (mci.dg)
2511                 cr14 |= MCCK_CR14_DEGRAD_SUB_MASK;
2512         if (mci.w)
2513                 cr14 |= MCCK_CR14_WARN_SUB_MASK;
2514
2515         mchk = mci.ck ? &inti.mchk : &irq.u.mchk;
2516         mchk->cr14 = cr14;
2517         mchk->mcic = mcck_info->mcic;
2518         mchk->ext_damage_code = mcck_info->ext_damage_code;
2519         mchk->failing_storage_address = mcck_info->failing_storage_address;
2520         if (mci.ck) {
2521                 /* Inject the floating machine check */
2522                 inti.type = KVM_S390_MCHK;
2523                 rc = __inject_vm(vcpu->kvm, &inti);
2524         } else {
2525                 /* Inject the machine check to specified vcpu */
2526                 irq.type = KVM_S390_MCHK;
2527                 rc = kvm_s390_inject_vcpu(vcpu, &irq);
2528         }
2529         WARN_ON_ONCE(rc);
2530 }
2531
2532 int kvm_set_routing_entry(struct kvm *kvm,
2533                           struct kvm_kernel_irq_routing_entry *e,
2534                           const struct kvm_irq_routing_entry *ue)
2535 {
2536         int ret;
2537
2538         switch (ue->type) {
2539         case KVM_IRQ_ROUTING_S390_ADAPTER:
2540                 e->set = set_adapter_int;
2541                 e->adapter.summary_addr = ue->u.adapter.summary_addr;
2542                 e->adapter.ind_addr = ue->u.adapter.ind_addr;
2543                 e->adapter.summary_offset = ue->u.adapter.summary_offset;
2544                 e->adapter.ind_offset = ue->u.adapter.ind_offset;
2545                 e->adapter.adapter_id = ue->u.adapter.adapter_id;
2546                 ret = 0;
2547                 break;
2548         default:
2549                 ret = -EINVAL;
2550         }
2551
2552         return ret;
2553 }
2554
2555 int kvm_set_msi(struct kvm_kernel_irq_routing_entry *e, struct kvm *kvm,
2556                 int irq_source_id, int level, bool line_status)
2557 {
2558         return -EINVAL;
2559 }
2560
2561 int kvm_s390_set_irq_state(struct kvm_vcpu *vcpu, void __user *irqstate, int len)
2562 {
2563         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2564         struct kvm_s390_irq *buf;
2565         int r = 0;
2566         int n;
2567
2568         buf = vmalloc(len);
2569         if (!buf)
2570                 return -ENOMEM;
2571
2572         if (copy_from_user((void *) buf, irqstate, len)) {
2573                 r = -EFAULT;
2574                 goto out_free;
2575         }
2576
2577         /*
2578          * Don't allow setting the interrupt state
2579          * when there are already interrupts pending
2580          */
2581         spin_lock(&li->lock);
2582         if (li->pending_irqs) {
2583                 r = -EBUSY;
2584                 goto out_unlock;
2585         }
2586
2587         for (n = 0; n < len / sizeof(*buf); n++) {
2588                 r = do_inject_vcpu(vcpu, &buf[n]);
2589                 if (r)
2590                         break;
2591         }
2592
2593 out_unlock:
2594         spin_unlock(&li->lock);
2595 out_free:
2596         vfree(buf);
2597
2598         return r;
2599 }
2600
2601 static void store_local_irq(struct kvm_s390_local_interrupt *li,
2602                             struct kvm_s390_irq *irq,
2603                             unsigned long irq_type)
2604 {
2605         switch (irq_type) {
2606         case IRQ_PEND_MCHK_EX:
2607         case IRQ_PEND_MCHK_REP:
2608                 irq->type = KVM_S390_MCHK;
2609                 irq->u.mchk = li->irq.mchk;
2610                 break;
2611         case IRQ_PEND_PROG:
2612                 irq->type = KVM_S390_PROGRAM_INT;
2613                 irq->u.pgm = li->irq.pgm;
2614                 break;
2615         case IRQ_PEND_PFAULT_INIT:
2616                 irq->type = KVM_S390_INT_PFAULT_INIT;
2617                 irq->u.ext = li->irq.ext;
2618                 break;
2619         case IRQ_PEND_EXT_EXTERNAL:
2620                 irq->type = KVM_S390_INT_EXTERNAL_CALL;
2621                 irq->u.extcall = li->irq.extcall;
2622                 break;
2623         case IRQ_PEND_EXT_CLOCK_COMP:
2624                 irq->type = KVM_S390_INT_CLOCK_COMP;
2625                 break;
2626         case IRQ_PEND_EXT_CPU_TIMER:
2627                 irq->type = KVM_S390_INT_CPU_TIMER;
2628                 break;
2629         case IRQ_PEND_SIGP_STOP:
2630                 irq->type = KVM_S390_SIGP_STOP;
2631                 irq->u.stop = li->irq.stop;
2632                 break;
2633         case IRQ_PEND_RESTART:
2634                 irq->type = KVM_S390_RESTART;
2635                 break;
2636         case IRQ_PEND_SET_PREFIX:
2637                 irq->type = KVM_S390_SIGP_SET_PREFIX;
2638                 irq->u.prefix = li->irq.prefix;
2639                 break;
2640         }
2641 }
2642
2643 int kvm_s390_get_irq_state(struct kvm_vcpu *vcpu, __u8 __user *buf, int len)
2644 {
2645         int scn;
2646         unsigned long sigp_emerg_pending[BITS_TO_LONGS(KVM_MAX_VCPUS)];
2647         struct kvm_s390_local_interrupt *li = &vcpu->arch.local_int;
2648         unsigned long pending_irqs;
2649         struct kvm_s390_irq irq;
2650         unsigned long irq_type;
2651         int cpuaddr;
2652         int n = 0;
2653
2654         spin_lock(&li->lock);
2655         pending_irqs = li->pending_irqs;
2656         memcpy(&sigp_emerg_pending, &li->sigp_emerg_pending,
2657                sizeof(sigp_emerg_pending));
2658         spin_unlock(&li->lock);
2659
2660         for_each_set_bit(irq_type, &pending_irqs, IRQ_PEND_COUNT) {
2661                 memset(&irq, 0, sizeof(irq));
2662                 if (irq_type == IRQ_PEND_EXT_EMERGENCY)
2663                         continue;
2664                 if (n + sizeof(irq) > len)
2665                         return -ENOBUFS;
2666                 store_local_irq(&vcpu->arch.local_int, &irq, irq_type);
2667                 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2668                         return -EFAULT;
2669                 n += sizeof(irq);
2670         }
2671
2672         if (test_bit(IRQ_PEND_EXT_EMERGENCY, &pending_irqs)) {
2673                 for_each_set_bit(cpuaddr, sigp_emerg_pending, KVM_MAX_VCPUS) {
2674                         memset(&irq, 0, sizeof(irq));
2675                         if (n + sizeof(irq) > len)
2676                                 return -ENOBUFS;
2677                         irq.type = KVM_S390_INT_EMERGENCY;
2678                         irq.u.emerg.code = cpuaddr;
2679                         if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2680                                 return -EFAULT;
2681                         n += sizeof(irq);
2682                 }
2683         }
2684
2685         if (sca_ext_call_pending(vcpu, &scn)) {
2686                 if (n + sizeof(irq) > len)
2687                         return -ENOBUFS;
2688                 memset(&irq, 0, sizeof(irq));
2689                 irq.type = KVM_S390_INT_EXTERNAL_CALL;
2690                 irq.u.extcall.code = scn;
2691                 if (copy_to_user(&buf[n], &irq, sizeof(irq)))
2692                         return -EFAULT;
2693                 n += sizeof(irq);
2694         }
2695
2696         return n;
2697 }